WEBVTT 1 00:00:10.300 --> 00:00:15.100 Regarding sample size and sampling methods. It is a common practice by 2 00:00:15.100 --> 00:00:19.100 us Indians that whenever we go to the grocery store or whenever 3 00:00:19.100 --> 00:00:22.700 we go to the vegetable or fruits market, before buying a whole 4 00:00:22.700 --> 00:00:27.800 lot of vegetables, we are tend to examine a 5 00:00:27.800 --> 00:00:32.800 sample of the whole lot of the things that we need to 6 00:00:32.800 --> 00:00:36.800 buy. Say for example, in this case, in this photograph, the person 7 00:00:36.800 --> 00:00:42.100 is trying to assess the quality of the wheat by taking 8 00:00:42.100 --> 00:00:46.800 few grains of wheat. Okay, so this is nothing but what is 9 00:00:47.100 --> 00:00:52.200 the concept of sampling in medical research. In medical research, what 10 00:00:52.200 --> 00:00:57.500 we want to do is we want to check the 11 00:00:57.500 --> 00:01:02.700 sample for doing medical research. Okay, whenever whenever we have 12 00:01:02.700 --> 00:01:06.700 a target population and we want to do a research related to 13 00:01:06.800 --> 00:01:11.800 any particular medical uh field. For example, say 14 00:01:11.900 --> 00:01:16.900 one doctor wants to conduct a research on the diabetes and 15 00:01:16.900 --> 00:01:21.600 say the foot ulcers in diabetes. Okay, so the target population will 16 00:01:21.600 --> 00:01:26.700 be the adult cases of diabetes and he 17 00:01:26.700 --> 00:01:32.600 wants to check how many diabetes cases they have foot 18 00:01:32.600 --> 00:01:38.300 ulcers. So what he will do? It is not necessary to target 19 00:01:38.300 --> 00:01:42.400 all the diabetes cases in his field practice area. Say for example, 20 00:01:42.400 --> 00:01:46.800 he is practicing in a city of Mumbai. So in the city 21 00:01:46.800 --> 00:01:50.700 of Mumbai, there will be thousands and lakhs of people who will 22 00:01:50.700 --> 00:01:54.300 be having diabetes. We just want to check how how much is 23 00:01:54.300 --> 00:01:58.500 the percentage of diabetes cases who are having foot ulcers. So what 24 00:01:58.500 --> 00:02:03.700 he will do? He will gather few of the patients from 25 00:02:03.700 --> 00:02:08.400 the Mumbai who are already diagnosed diabetic and he will conduct his 26 00:02:08.400 --> 00:02:13.800 research. he or she will conduct his the research on this few 27 00:02:13.800 --> 00:02:18.600 number of cases of diabetes. So this is the concept of sampling. 28 00:02:18.700 --> 00:02:24.400 In sampling, we want to do a particular research regarding a particular 29 00:02:24.500 --> 00:02:28.600 medical topic or any medical or social or any topic that we 30 00:02:28.600 --> 00:02:32.400 want to do research in that community, in that population. It is 31 00:02:32.400 --> 00:02:37.100 not necessary, it is not important to include all the patients or 32 00:02:37.100 --> 00:02:42.800 all the participants from the population. We can just take few of 33 00:02:42.800 --> 00:02:46.800 them and we conduct the research. Sampling, it is a scientific method 34 00:02:48.200 --> 00:02:53.500 by which some members of the population are selected as representatives 35 00:02:53.500 --> 00:02:57.900 of the entire population. It is a significant stage of scientific work 36 00:02:58.200 --> 00:03:02.300 and if ignored, the results may become invalid and futile. So it 37 00:03:02.300 --> 00:03:06.900 is one of the core part of research of in the medical 38 00:03:06.900 --> 00:03:12.100 field. So if we avoid scientific sampling method and if we 39 00:03:12.400 --> 00:03:16.600 fail to mention how the sample was sample size was calculated and 40 00:03:16.600 --> 00:03:20.500 how we came to a conclusion for this research that this much 41 00:03:20.500 --> 00:03:25.800 sample is sufficient, then it may be invalid and all 42 00:03:25.800 --> 00:03:29.800 the research may go in vain. The common terms in the sample 43 00:03:30.000 --> 00:03:35.400 sampling technique is the population. Population is the entire set of individual 44 00:03:35.400 --> 00:03:40.700 or objects having some common characteristics. Finite population, all units 45 00:03:40.700 --> 00:03:45.900 can be counted before sampling. Infinite all units cannot be enlisted. 46 00:03:47.900 --> 00:03:52.900 Now population is the whole method in which the sampling sample 47 00:03:52.900 --> 00:03:58.200 size can be there. Sampling is a small section or 48 00:03:58.200 --> 00:04:04.200 a subset of population selected to participate in the study. Now 49 00:04:05.000 --> 00:04:09.000 what is mean by sampling? Sample is a small section and sampling 50 00:04:09.000 --> 00:04:13.100 is the process of selecting subjects who are representative of the population. 51 00:04:13.600 --> 00:04:19.000 That is the difference between sample and sampling. Sample is the exact 52 00:04:19.000 --> 00:04:24.200 number of cases, exact number of participants that we will be taking 53 00:04:24.200 --> 00:04:29.200 into our study while sampling is the process of selecting these cases 54 00:04:29.300 --> 00:04:34.600 or process of selecting these samples. So here we have the target 55 00:04:34.600 --> 00:04:39.800 population in which that study will be done. Here we have 56 00:04:39.900 --> 00:04:45.000 study population which will be accessible. Study population is the population 57 00:04:45.000 --> 00:04:49.300 that will be accessible, will be available to be included in the 58 00:04:49.300 --> 00:04:54.400 study and from this study population, we will include this sample, 59 00:04:55.000 --> 00:04:59.100 these few cases in our research. Okay, so that is the basic 60 00:04:59.500 --> 00:05:03.900 concept behind selecting the number of cases, number of participants for a 61 00:05:03.900 --> 00:05:09.200 research in the sampling. So the target population is 62 00:05:09.200 --> 00:05:13.800 the population where we want to generalize our finding. Say for example, 63 00:05:13.800 --> 00:05:18.800 in that diabetes foot ulcer study in the city of Mumbai, the 64 00:05:18.800 --> 00:05:23.400 doctor finds out that out of the 200 cases of diabetes studied, 65 00:05:23.500 --> 00:05:28.600 there were 40 cases of foot ulcers. That means the foot ulcer 66 00:05:28.600 --> 00:05:34.000 percentage is 20% in the diabetic cases. And this 20% 67 00:05:34.400 --> 00:05:39.400 prevalence of foot ulcer in diabetic, it can be generalized in the 68 00:05:39.400 --> 00:05:43.800 target population of the all the diabetes cases in the city of 69 00:05:43.800 --> 00:05:48.900 Mumbai. So we can say that in the city of Mumbai, there 70 00:05:48.900 --> 00:05:54.500 is a prevalence of 20% of diabetic foot ulcers in 71 00:05:54.500 --> 00:05:59.200 the cases of diabetes. Even though we have just studied 200 cases 72 00:05:59.200 --> 00:06:02.400 of diabetes, there can be thousands and lakhs of cases in the 73 00:06:02.400 --> 00:06:07.600 city of Mumbai. Similarly, we can generalize this on even larger 74 00:06:07.600 --> 00:06:12.700 scale. The study population is the population that you get access to 75 00:06:12.800 --> 00:06:18.000 and the sample is the exact number of cases, the person 76 00:06:18.000 --> 00:06:21.700 that we have conducted the study on. Study population depends on the 77 00:06:21.700 --> 00:06:26.900 research question, how many injections do people receive each year in India? 78 00:06:26.900 --> 00:06:30.600 Okay, so what will be the study population in this case? It 79 00:06:30.600 --> 00:06:35.600 will be the population of India. If the study population we 80 00:06:35.600 --> 00:06:40.800 need to see how many needle stick healthcare workers experience each 81 00:06:40.800 --> 00:06:44.900 year in India. So what are the needle stick injuries? Say for 82 00:06:44.900 --> 00:06:50.900 example, a nurse is collecting blood sample of a patient. Accidentally, 83 00:06:50.900 --> 00:06:56.300 what happens is after collecting the blood sample, the needle gets pricked 84 00:06:56.400 --> 00:07:00.300 to the hand of the nurse. So what is this? This is 85 00:07:00.300 --> 00:07:04.300 a needle stick injury. So we want to know how many needle 86 00:07:04.300 --> 00:07:09.600 stick injuries are there to the healthcare workers each 87 00:07:09.600 --> 00:07:13.700 year. So the study population will be healthcare workers of India. If 88 00:07:13.700 --> 00:07:16.800 we want to know how many hospitals have a needle stick prevention 89 00:07:16.800 --> 00:07:20.300 policy in India, what will be the study population? It will be 90 00:07:20.300 --> 00:07:25.000 the number of hospitals in India. So this is the way we 91 00:07:25.100 --> 00:07:30.100 from the research question, we can decide what will be our study 92 00:07:30.100 --> 00:07:35.100 population. Why sampling is required? It is required to obtain information from 93 00:07:35.100 --> 00:07:40.700 large populations and to ensure the efficiency of a study. To 94 00:07:40.700 --> 00:07:46.000 obtain more accurate information, we we do the sampling. The 95 00:07:46.000 --> 00:07:50.600 sampling terms are sampling unit or basic sampling unit which is the 96 00:07:50.600 --> 00:07:55.000 elementary unit that will be sampled. Okay, so this sampling unit in 97 00:07:55.000 --> 00:08:00.200 the study where we wanted to know how many hospitals have 98 00:08:00.200 --> 00:08:04.600 the needle stick injury policy. Here the sampling unit will be a 99 00:08:04.600 --> 00:08:09.900 single hospital. Okay, a single hospital will be the elementary unit. 100 00:08:10.200 --> 00:08:13.100 It will not be any patient in that hospital or any doctor 101 00:08:13.100 --> 00:08:17.400 in that hospital because our research question is how many hospitals have 102 00:08:17.500 --> 00:08:23.100 needle stick injury policy. So here elementary unit will be the one 103 00:08:23.100 --> 00:08:28.300 single hospital. In the in the study of the diabetic foot foot 104 00:08:28.300 --> 00:08:32.500 ulcers in the city of Mumbai, elementary unit of the research will 105 00:08:32.500 --> 00:08:36.800 be one patient of diabetes in the state in the city of 106 00:08:36.800 --> 00:08:41.100 Mumbai. Okay. So these are these can be people, these can be 107 00:08:41.100 --> 00:08:44.900 healthcare workers, these can be hospitals. Okay, so as I have explained 108 00:08:45.600 --> 00:08:48.800 what is the sampling frame? It is the list of all sampling 109 00:08:48.800 --> 00:08:53.800 units in the population. The sampling scheme, it is the method used 110 00:08:53.800 --> 00:08:57.200 to select sampling units from the sampling frame. So now we know 111 00:08:57.200 --> 00:09:01.500 what is sampling unit. Sampling frame, we need to have a list 112 00:09:01.500 --> 00:09:04.900 of all the sampling units in the population. If we want to 113 00:09:04.900 --> 00:09:09.000 study the number of hospitals with needle stick injury policy, we need 114 00:09:09.000 --> 00:09:12.600 to have a list of all the hospitals in that particular area 115 00:09:12.600 --> 00:09:15.900 where we want to conduct our study. Say for example, in India, 116 00:09:16.000 --> 00:09:20.500 we need to have list of all the hospitals in India. If 117 00:09:20.500 --> 00:09:25.100 we want to study the diabetic foot foot ulcer, we need to 118 00:09:25.100 --> 00:09:29.200 have a list of all the diabetes cases in the city of 119 00:09:29.200 --> 00:09:33.800 Mumbai. Sampling scheme is the method used to select sampling units from 120 00:09:33.800 --> 00:09:38.400 the sampling frame. So sampling scheme is how exactly we are going 121 00:09:38.400 --> 00:09:43.800 to select those 200 diabetes cases from the city of Mumbai who 122 00:09:43.800 --> 00:09:48.200 who may have thousands and lakhs of diabetes patients. Okay. So how 123 00:09:48.200 --> 00:09:53.600 exactly are we going to select this diabetes cases? Is 124 00:09:53.600 --> 00:09:59.400 the sample representative one in terms of time, place and person? What 125 00:09:59.400 --> 00:10:04.400 does it mean representative? That means the 200 cases that 126 00:10:04.400 --> 00:10:09.400 we have chosen in our study for diabetes, it should be similar 127 00:10:09.400 --> 00:10:14.800 to all the say for example, there are 200,000 cases of diabetes 128 00:10:14.800 --> 00:10:18.400 in the study of in the city of Mumbai. Out of 200,000 129 00:10:18.400 --> 00:10:23.800 diabetes cases, we have we have taken the 200 cases of diabetes 130 00:10:23.800 --> 00:10:29.400 for our research. So we are taking 0.1% of 131 00:10:29.400 --> 00:10:34.600 the whole number of diabetes cases. 200,000 200 divided by 200,000 132 00:10:34.600 --> 00:10:39.600 into 100 taken into percentage is 0.1 133 00:10:39.700 --> 00:10:45.000 percent. Okay. So in terms of time, whether all 134 00:10:45.000 --> 00:10:50.200 have diabetes for a similar duration, whether all are residing in 135 00:10:50.500 --> 00:10:55.800 the urban like urban slums or in the high rise area 136 00:10:55.800 --> 00:11:00.600 or in the high income group areas. And in the terms of 137 00:11:01.400 --> 00:11:06.900 person, whether the male and female distribution, age distribution is similar in 138 00:11:06.900 --> 00:11:10.200 the 200 cases that we are going to study as compared to 139 00:11:10.200 --> 00:11:14.600 the 200,000 population. So it is the whether the sample is representative 140 00:11:14.600 --> 00:11:18.800 one. For example, the Ministry of Health of a country X wants 141 00:11:18.800 --> 00:11:23.000 to estimate the proportion of children in elementary schools who have been 142 00:11:23.000 --> 00:11:28.000 immunized against childhood infectious diseases. So we have various childhood infectious 143 00:11:28.000 --> 00:11:33.000 diseases and in a country the Ministry of Health wants 144 00:11:33.000 --> 00:11:38.300 to see how many percentage of children in elementary schools are 145 00:11:38.300 --> 00:11:43.200 immunized. The task must be completed in a month. We have only 146 00:11:43.200 --> 00:11:46.900 one month. The objective is to estimate the proportion of immunized children. 147 00:11:47.900 --> 00:11:53.000 The types of sampling. So here what is the challenge, biggest challenge 148 00:11:53.100 --> 00:11:58.100 is in just one month, we cannot see the 149 00:11:58.100 --> 00:12:03.200 whole country. The whole country will have lakhs and crores of 150 00:12:03.200 --> 00:12:08.100 children going to the elementary school. but in one month, we can 151 00:12:08.100 --> 00:12:12.400 just do a few number of schools and we can check few 152 00:12:12.400 --> 00:12:17.000 number of children only. The types of sampling are broadly classified into 153 00:12:17.000 --> 00:12:22.200 probability sampling and non-probability sampling. Probability sampling we have simple random 154 00:12:22.200 --> 00:12:27.400 sampling, stratified random sampling, systematic sampling and cluster sampling, 155 00:12:27.500 --> 00:12:33.200 multi-phase and multi-stage sampling. Non-probability sampling we have purposive, quota, 156 00:12:33.300 --> 00:12:38.300 snowball sampling, sequential sampling, convenience sampling. Okay. So this is 157 00:12:38.300 --> 00:12:43.300 the basic difference between probability and non-probability sampling. As the 158 00:12:43.300 --> 00:12:48.600 name suggests, probability sampling it has got a every 159 00:12:50.500 --> 00:12:55.700 person, every individual, every unit, every sampling unit in the sampling 160 00:12:55.700 --> 00:13:00.800 frame will have a nth probability 161 00:13:00.800 --> 00:13:05.000 of getting selected in the probability sampling. While non-probability sampling, there is 162 00:13:05.000 --> 00:13:10.200 no any such probability that there is no any particular 163 00:13:10.200 --> 00:13:15.100 formula or particular figure where the samples will be selected. Okay, so 164 00:13:15.100 --> 00:13:19.200 that is the basic difference. What is probability sampling? Sampling method where 165 00:13:19.300 --> 00:13:23.200 every member of the population has an equal chance of being selected 166 00:13:23.200 --> 00:13:28.100 as a sample. Only sampling method that allows to draw valid conclusions 167 00:13:28.100 --> 00:13:33.300 about a population. Random sampling which is the first in 168 00:13:33.300 --> 00:13:37.800 case of probability sampling. It removes possibility of bias in selection of 169 00:13:37.800 --> 00:13:42.600 subjects. Okay, what is bias? Bias is the mistake that can occur 170 00:13:42.700 --> 00:13:47.400 in the selection of subjects. Random sampling ensures that each subject has 171 00:13:47.400 --> 00:13:52.200 a known probability of being chosen and it allows application of statistical 172 00:13:52.200 --> 00:13:56.700 theory. The sampling error no sample is perfect mirror image of the 173 00:13:56.700 --> 00:14:01.800 population. Magnitude of error can be measured in probability samples expressed 174 00:14:01.800 --> 00:14:06.200 as a standard error of mean proportions or differences and functions of 175 00:14:06.200 --> 00:14:10.300 sample size and variability in measurement can be done by sampling error. 176 00:14:11.100 --> 00:14:15.100 What is simple random sampling principle? There is equal chance for each 177 00:14:15.100 --> 00:14:19.900 sampling unit procedure. First, we need to number all the units and 178 00:14:19.900 --> 00:14:23.600 then we need to randomly draw units. Okay, so we have 200,000 179 00:14:23.600 --> 00:14:27.300 cases of diabetes in the city of Mumbai. What we do is 180 00:14:27.400 --> 00:14:32.700 we enlist and we number all the 200,000 cases of diabetes from 181 00:14:32.700 --> 00:14:37.300 1 to 200,000 and if our sample size is 200, what we 182 00:14:37.300 --> 00:14:42.400 do is we randomly draw the 200 numbers in from 183 00:14:42.400 --> 00:14:47.600 these 200,000 samples. Okay. Advantage it is a simple and sampling 184 00:14:47.600 --> 00:14:52.500 either is easily measured. Disadvantages we need complete list of units and 185 00:14:52.500 --> 00:14:57.600 it does not achieve best representation. Okay. So there is a a 186 00:14:57.600 --> 00:15:02.800 simple random sampling and the random numbers can be obtained 187 00:15:02.800 --> 00:15:07.800 from the biostatistics books, research methodology books where the random numbers list 188 00:15:07.800 --> 00:15:12.800 will be provided or we can just go to the software for 189 00:15:13.500 --> 00:15:18.700 biostatistical research, medical research which is EP info and 190 00:15:18.700 --> 00:15:23.500 the website calculations are also available online on the EP info. So 191 00:15:24.000 --> 00:15:27.200 after the end of this lecture, we will see how we can 192 00:15:27.300 --> 00:15:32.300 get the random number list from the EP info software and 193 00:15:32.300 --> 00:15:37.400 we will also do one sample size calculation in the software. 194 00:15:37.600 --> 00:15:41.100 Okay, so this is the example of simple random sampling. Here we 195 00:15:41.100 --> 00:15:45.400 have 1 to 48 list and out of these 1 to 48 196 00:15:45.400 --> 00:15:50.600 randomly we are selecting these four cases. So number nine Dennis, number 197 00:15:50.600 --> 00:15:55.700 18 Frank, number 32 Mike, number 40 LM we are selecting 198 00:15:55.700 --> 00:16:00.700 in our sample. What is systematic sampling? A unit drawn every 199 00:16:00.700 --> 00:16:06.100 K unit. So here we have a number which repeats and 200 00:16:06.100 --> 00:16:11.200 we take every that number. Equal chance of being drawn 201 00:16:11.400 --> 00:16:16.200 from each unit procedure. First, we need to calculate the sampling interval. 202 00:16:16.200 --> 00:16:20.000 K is equal to N by N. So what is this sampling 203 00:16:20.000 --> 00:16:24.400 interval? For example, in the diabetes studies, we have 200,000 diabetes cases. 204 00:16:25.300 --> 00:16:29.200 Capital N is the whole population. It is 200,000. Small N is 205 00:16:29.200 --> 00:16:33.800 our sample size. It is 200. So 200,000 by 200 that will 206 00:16:33.800 --> 00:16:39.000 be 1,000. We need to see a random number of less than 207 00:16:39.100 --> 00:16:42.700 K that is less than 1,000 for starting. Say for example, our 208 00:16:42.700 --> 00:16:48.100 random number is 125. So this is a random number 125. 209 00:16:48.500 --> 00:16:52.700 Every K unit from the first unit, we can select and we 210 00:16:52.700 --> 00:16:58.500 can prepare a sample. So 125 plus 211 00:16:59.400 --> 00:17:04.600 the first patient will be 125 number. Second patient will be 212 00:17:04.800 --> 00:17:10.100 1,125. Third patient will be 2,125. Fourth 213 00:17:10.100 --> 00:17:15.200 number will be 3,125 because K or the sampling interval 214 00:17:15.200 --> 00:17:20.300 here is 1,000. That means out of 200,000 patients, we need 215 00:17:20.300 --> 00:17:25.100 to have 200 diabetes cases and for that sampling interval of 1,000 216 00:17:25.100 --> 00:17:30.600 is there. When we have selected a random number of 125 after 217 00:17:30.600 --> 00:17:34.800 125 which which is the first case, first patient to be included 218 00:17:34.800 --> 00:17:39.800 in the study. Every 1,000 case we will be adding to our 219 00:17:39.800 --> 00:17:45.200 research. Advantages ensures representativeness across list and easy to implement. 220 00:17:45.200 --> 00:17:50.500 Disadvantage dangerous if the list has got cycles. Example of systematic sampling 221 00:17:50.500 --> 00:17:55.400 if we want to study households which have got supply of clean 222 00:17:55.400 --> 00:18:01.000 drinking water. We have selected the a colony 223 00:18:01.100 --> 00:18:05.500 and we have found our sampling interval which is eight house or 224 00:18:05.500 --> 00:18:09.800 the sampling interval is eight. We started with the third house or 225 00:18:09.800 --> 00:18:13.800 the number three house and we went to the three plus eight. 226 00:18:13.800 --> 00:18:17.500 1 2 3 4 5 6 7 8. This is the 11th 227 00:18:17.500 --> 00:18:22.000 house. So this is the sampling interval eight 11th house we will 228 00:18:22.000 --> 00:18:27.200 be adding. Again, we will go eight house one 1 229 00:18:27.200 --> 00:18:31.000 2 3 4 5 6 7 8. 19th house we will add 230 00:18:31.100 --> 00:18:36.300 again the 27th house, then 35th house, 231 00:18:36.400 --> 00:18:41.800 then 43rd house, then 51st house. So upper like 232 00:18:41.800 --> 00:18:46.400 this every eighth house we will be adding in our research. Now 233 00:18:46.400 --> 00:18:51.300 what is stratified sampling? So this was systematic random sampling. Okay. This 234 00:18:51.300 --> 00:18:55.300 is now stratified sampling. What is stratified sampling? Principle is to classify 235 00:18:55.300 --> 00:19:00.600 population into homogeneous groups like strata. Draw a sample in each strata 236 00:19:00.600 --> 00:19:04.100 and combine these results of all strata. Advantages it is a more 237 00:19:04.100 --> 00:19:09.800 precise if variable associated with the strata. All subgroups are represented 238 00:19:10.000 --> 00:19:14.900 allowing separate conclusions about each of them. Disadvantage sampling error is difficult 239 00:19:14.900 --> 00:19:19.900 to measure. Loss of precision if small numbers sampled in individual 240 00:19:19.900 --> 00:19:25.000 strata. Example of stratified sampling are to estimate vaccination 241 00:19:25.000 --> 00:19:29.900 coverage in a country. One sample drawn in each region and then 242 00:19:29.900 --> 00:19:34.900 estimate calculated for each strata. Okay, so we need to divide the 243 00:19:35.600 --> 00:19:40.600 sample into different strata. So for vaccination coverage 244 00:19:40.600 --> 00:19:45.300 in the country, we have to draw one sample from each region 245 00:19:45.300 --> 00:19:50.800 in region of Northern India, Western India, Eastern India, Southern India, Northeast 246 00:19:51.400 --> 00:19:56.700 states. So we have divided five regions and we take samples 247 00:19:56.700 --> 00:20:01.700 from each region and we calculate the estimate from 248 00:20:01.700 --> 00:20:06.800 each strata. Each strata weighed into obtain estimate for 249 00:20:06.800 --> 00:20:11.000 a country. Now what is cluster sampling? Principle is the random samples 250 00:20:11.000 --> 00:20:15.100 of groups or clusters of units are to be included in the 251 00:20:15.100 --> 00:20:20.400 study. All or proportion of units are included from the selected clusters. 252 00:20:20.700 --> 00:20:24.500 Advantage it is simple, no list of units required and less travel 253 00:20:24.500 --> 00:20:27.600 or resources are required. Let me give you an example of a 254 00:20:27.600 --> 00:20:32.800 cluster sampling for our Indra Dhanush project that we have done a 255 00:20:32.800 --> 00:20:36.800 few years back. What we needed to do is we need we 256 00:20:36.800 --> 00:20:42.200 needed to check the immunization status of a district. What we 257 00:20:42.400 --> 00:20:48.200 did was we made a clusters of seven villages 258 00:20:48.200 --> 00:20:53.200 in that district and from the each cluster we selected 30 cases 259 00:20:53.200 --> 00:20:58.000 of children under five years of age. So what what was the 260 00:20:58.000 --> 00:21:02.600 advantage? We didn't require any list of all the children of all 261 00:21:02.600 --> 00:21:07.700 the villages. We just selected seven random villages and we went into 262 00:21:07.700 --> 00:21:11.500 those villages and we selected 30 random children of less than five 263 00:21:11.500 --> 00:21:15.100 years of age from each village. So this is a cluster sampling 264 00:21:15.100 --> 00:21:19.500 where we are taking the clusters of seven different villages and in 265 00:21:19.500 --> 00:21:24.800 that one cluster we are taking 30 cases. Disadvantages imprecise 266 00:21:24.800 --> 00:21:30.000 if clusters are homogeneous and sampling either is difficult to measure. The 267 00:21:30.000 --> 00:21:33.300 sampling unit is not a subject but a group of subjects. So 268 00:21:33.300 --> 00:21:38.500 that every group of 30 children under five from the villages was 269 00:21:38.700 --> 00:21:42.400 the sampling unit. It is assumed that the variability among clusters is 270 00:21:42.400 --> 00:21:46.300 minimal. The variability within each cluster is what is observed in the 271 00:21:46.300 --> 00:21:51.600 general population. Two stages of cluster sampling first stage is probability proportional 272 00:21:51.600 --> 00:21:55.700 to the size. Select the number of clusters to be included. Compute 273 00:21:55.700 --> 00:21:59.200 a cumulative list of the populations in in each unit with a 274 00:21:59.200 --> 00:22:02.600 grand total. Divide the grand total by the number of clusters and 275 00:22:02.600 --> 00:22:08.000 obtain the sampling interval. So how exactly we come to know 276 00:22:08.000 --> 00:22:11.100 about how many villages are to be included? For that, we need 277 00:22:11.100 --> 00:22:15.600 to have the number of clusters, the total number of villages that 278 00:22:15.600 --> 00:22:20.300 will be available in that district. Say for example, there were 140 279 00:22:20.300 --> 00:22:25.300 villages in that district. And we we got 280 00:22:25.400 --> 00:22:30.400 the sampling interval of 20. So for every 281 00:22:30.700 --> 00:22:35.400 one village, for every 20 villages, we had to study one village 282 00:22:35.400 --> 00:22:40.400 for that particular study. The sampling interval being 20 and total number 283 00:22:40.400 --> 00:22:44.900 of villages in the district was 140. So we divided 140 by 284 00:22:44.900 --> 00:22:49.300 the sampling interval 20. So we got a cluster sample of seven. 285 00:22:49.400 --> 00:22:54.800 Okay. Choose a random number and identify the plus first cluster. So 286 00:22:54.900 --> 00:23:00.700 we name all the number, all the villages 1 to 140 287 00:23:00.800 --> 00:23:05.300 and we start with the third village and every seventh every 20th 288 00:23:05.300 --> 00:23:10.500 village we go and we take the sample. So third, 23rd, 289 00:23:10.700 --> 00:23:16.000 43rd, 63rd, 83rd like that we chose seven 290 00:23:16.000 --> 00:23:20.700 villages. Add the sampling interval and identify the second cluster. By repeating 291 00:23:20.700 --> 00:23:24.800 the same procedure, identify all the clusters. And the second stage is 292 00:23:24.800 --> 00:23:28.100 in each cluster select a random sample using a sampling frame of 293 00:23:28.100 --> 00:23:33.100 subjects or households. So same way in the village also we can 294 00:23:33.100 --> 00:23:38.100 select by different random number sample. You can select those 295 00:23:38.100 --> 00:23:43.200 30 cases each. This was all regarding the cluster sampling. Now what 296 00:23:43.200 --> 00:23:47.500 is multi-stage sampling? As the name suggests, it has got stages. So 297 00:23:47.500 --> 00:23:52.700 this is the stage one some part of those who are in 298 00:23:52.700 --> 00:23:56.100 the stage one will go to the stage two. Some part of 299 00:23:56.100 --> 00:23:58.600 those who are in the stage two will go to the stage 300 00:23:58.600 --> 00:24:04.200 three. So this is a stage wise sampling. Principle is several 301 00:24:04.200 --> 00:24:09.800 chained samples are required and several statistical units are required. Advantage is 302 00:24:10.000 --> 00:24:13.700 no complete list of population is required and more feasible approach for 303 00:24:13.700 --> 00:24:19.100 large population. Disadvantage several sampling lists we have to make and sampling 304 00:24:19.100 --> 00:24:24.500 error is difficult to measure. So this multi-stage camp sampling 305 00:24:24.900 --> 00:24:30.100 we from India, we choose three states 1 2 3 and 306 00:24:30.100 --> 00:24:35.100 we exclude these two states and from this one state we choose 307 00:24:35.100 --> 00:24:38.900 three districts 1 to 3 and from these three districts we choose 308 00:24:38.900 --> 00:24:43.900 one district and we include these three villages from that district from 309 00:24:43.900 --> 00:24:47.500 the whole country of India. So this is from the India we 310 00:24:47.500 --> 00:24:52.500 choose. So out of total 27 villages we have chose 311 00:24:52.500 --> 00:24:56.700 these three villages from these nine districts from these total three states 312 00:24:56.800 --> 00:24:59.900 in the country of India. So this is the way we go 313 00:25:00.330 --> 00:25:04.830 in the multi-stage sampling. Multi-phase sampling, sampling is done in different phases. 314 00:25:05.030 --> 00:25:09.030 Part of information is collected from whole sample and part of information 315 00:25:09.030 --> 00:25:14.030 from sub-sample. Here each sample is adequately studied before another sample is 316 00:25:14.030 --> 00:25:19.230 drawn from it. Example a tuberculosis study. Tuberculosis multi-phase 317 00:25:19.230 --> 00:25:24.230 study phase one first Mantoux test was done to all 318 00:25:24.230 --> 00:25:28.830 the cases who are suspected of tuberculosis. Out of those who are 319 00:25:28.830 --> 00:25:33.830 Mantoux positive, we will conduct X-ray, chest X-ray. 320 00:25:34.230 --> 00:25:38.730 So what will happen? Mantoux test will be positive in suspected cases 321 00:25:38.730 --> 00:25:43.530 of tuberculosis. Those who are negative Mantoux test, we will not add 322 00:25:43.530 --> 00:25:46.730 them in the phase two. We will add only those who are 323 00:25:47.330 --> 00:25:50.930 Mantoux test positive for the phase two. In the phase two, we 324 00:25:50.930 --> 00:25:56.030 will do their chest X-ray. For phase three, those who are chest 325 00:25:56.030 --> 00:26:01.630 X-ray positive only we will go for the sputum examination. 326 00:26:01.630 --> 00:26:06.130 It will be either sputum microscopy or it will be the CBNAAT 327 00:26:06.230 --> 00:26:11.230 assessment of the sputum which is the cartridge-based nucleic acid amplification test. 328 00:26:12.030 --> 00:26:16.930 So this is the phase-wise sampling. Those who need sputum examination will 329 00:26:16.930 --> 00:26:20.930 be small as compared to to all the cases that we have 330 00:26:20.930 --> 00:26:26.530 studied by doing the Mantoux test, okay? So the difference between multiplex 331 00:26:26.530 --> 00:26:31.530 and multi-stage is multi-stage sampling frame is different at each state while 332 00:26:31.830 --> 00:26:35.930 multi-phase is samples are drawn from the same phase. It is used 333 00:26:35.930 --> 00:26:40.230 when insufficient budget to collect data from the whole sample is available. 334 00:26:41.230 --> 00:26:44.230 So this was all about the probability sampling. Now coming to the 335 00:26:44.230 --> 00:26:49.230 non-probability sampling, please remember for hospital-based studies, 336 00:26:49.330 --> 00:26:54.430 the most common type of sampling method used are non-probability sampling. For 337 00:26:54.430 --> 00:26:59.730 community studies, for epidemiological studies, for field research, it is very important 338 00:26:59.730 --> 00:27:05.030 that the type of sampling method used is probability sampling. A community-based 339 00:27:05.030 --> 00:27:09.830 study which is being which is used which is done using non-probability 340 00:27:09.830 --> 00:27:14.830 sampling is not a very good type of medical research. So for 341 00:27:14.930 --> 00:27:19.930 community-based studies, probability sampling are the best method of sampling to be 342 00:27:19.930 --> 00:27:25.330 used and for hospital-based studies, we can use non-probability sampling or 343 00:27:25.430 --> 00:27:30.030 for few of the particular studies also we use non-probability sampling. We 344 00:27:30.030 --> 00:27:34.630 will just go through them. It is a non-scientific sampling method. Every 345 00:27:34.630 --> 00:27:37.530 member of the population is not given a chance to be part 346 00:27:37.530 --> 00:27:41.530 of the sample. Choice depends on judgment of the researcher. What is 347 00:27:41.530 --> 00:27:47.130 judgment sampling? It is a purposive or deliberate sampling. Quota 348 00:27:47.130 --> 00:27:51.630 sampling, selecting 10 patients from each port. So quota sampling is we 349 00:27:51.630 --> 00:27:56.030 need to see what is the patient satisfaction regarding the services provided 350 00:27:56.030 --> 00:27:59.830 in the hospital. So for that, what we do is we select 351 00:27:59.830 --> 00:28:03.630 10 cases from each ward, 10 from the medicine male ward, 10 352 00:28:03.630 --> 00:28:07.130 from the medicine female ward, 10 from the gynecology ward, 10 from 353 00:28:07.130 --> 00:28:11.130 the surgery male, surgery female, 10 from the ortho male, orthopedic female, 354 00:28:11.330 --> 00:28:15.330 10 from the pediatric male and female. So we have this quota 355 00:28:15.330 --> 00:28:20.630 sampling. This is stratified sampling minus randomization. Sequential 356 00:28:20.630 --> 00:28:24.630 sampling, it is a consecutive sampling. Number of samples are drawn one 357 00:28:24.630 --> 00:28:28.630 after another from the population. We need to study what are the 358 00:28:29.330 --> 00:28:43.230 complications in diabetes. So so 359 00:28:43.330 --> 00:28:46.830 sequential sampling, what we do? The complications in the cases of diabetes 360 00:28:46.830 --> 00:28:49.630 in a hospital we want to study. What we will do? We 361 00:28:49.630 --> 00:28:54.730 will we will take the patients patients of diabetes as they come 362 00:28:54.730 --> 00:28:59.130 in the hospital one hour after the another. Now what is snowball 363 00:28:59.130 --> 00:29:03.930 sampling? Initial sample is selected by probability method and additional samples are 364 00:29:03.930 --> 00:29:08.730 obtained by referral from the initial respondent. So snowball sampling is a 365 00:29:08.730 --> 00:29:13.730 first a small proportion of samples we select by probability 366 00:29:13.730 --> 00:29:19.230 method and because of the rare form of these cases 367 00:29:19.230 --> 00:29:23.830 or the difficult to get the samples, we what we do is 368 00:29:23.830 --> 00:29:29.830 we refer the other cases to known cases from this first 369 00:29:29.830 --> 00:29:34.530 initial sample, okay? So I remember a snowball sampling method that we 370 00:29:34.530 --> 00:29:39.530 had done for the adolescent boys who are involved in 371 00:29:39.630 --> 00:29:43.430 the intravenous drug abuse in the urban slums of Mumbai. So that 372 00:29:43.430 --> 00:29:46.930 was one of the researches that we carried out and what we 373 00:29:46.930 --> 00:29:51.330 did is we identified four or five boys by the probability sampling 374 00:29:51.330 --> 00:29:56.630 method who were known intravenous drug abusers and 375 00:29:56.730 --> 00:30:01.830 we took them into confidence. We told them that no 376 00:30:01.830 --> 00:30:04.830 any legal action will be taken. We are just doctors. We want 377 00:30:04.830 --> 00:30:08.030 to inquire why you are all behind this and we want to 378 00:30:08.030 --> 00:30:12.230 help you, okay? So then they opened up and then they told 379 00:30:12.230 --> 00:30:15.730 okay, so I have two friends living in this jhopdi. I have 380 00:30:15.730 --> 00:30:20.930 two friends living in this gully. They also are habituated for 381 00:30:20.930 --> 00:30:25.730 the intravenous drug abuse. So this is how the number of samples 382 00:30:25.730 --> 00:30:31.230 by this referral from these close circuits went on increasing 383 00:30:31.230 --> 00:30:35.730 and this is how we conducted this snowball sampling. It is useful 384 00:30:35.730 --> 00:30:40.730 in locating rare cases. Then we have census sampling or universe sampling. 385 00:30:40.830 --> 00:30:44.830 A study using all available elements or members of a population. Say 386 00:30:44.830 --> 00:30:50.230 for example, we need to know what is the average 387 00:30:50.530 --> 00:30:56.230 height of the medical students admitted to a medical 388 00:30:56.230 --> 00:31:00.530 college. So what we will do? We will include all the MBBS 389 00:31:00.530 --> 00:31:04.130 students who have been admitted in a particular medical college in a 390 00:31:04.130 --> 00:31:08.930 city. Say for example, there are 200 per batch and there are 391 00:31:08.930 --> 00:31:13.030 four MBBS batches and one interns batch. So we have 1,000 medical 392 00:31:13.030 --> 00:31:16.430 students in that medical college. So what we will do? We will 393 00:31:16.530 --> 00:31:20.830 use universal method of sampling and we will include all the 1,000 394 00:31:20.830 --> 00:31:25.630 cases, all the 1,000 medical students for checking their mean height. When 395 00:31:25.630 --> 00:31:30.630 your population is very small, when you have extensive resources and 396 00:31:30.630 --> 00:31:34.030 when you don't expect a very high response. So what will you 397 00:31:34.030 --> 00:31:38.430 do? We include all of them in the study. Convenience sampling, what 398 00:31:38.430 --> 00:31:42.430 we mostly use in our studies? The samples are included as per 399 00:31:42.430 --> 00:31:48.730 the convenience of the researcher and so the merits of probability 400 00:31:49.130 --> 00:31:53.830 the convenience sampling are it is easy to obtain the sample and 401 00:31:53.830 --> 00:31:57.830 effective in testing the instruments and used in qualitative studies and demerits 402 00:31:57.830 --> 00:32:02.930 of non-probability sampling are they lack generalization, no representativeness and there 403 00:32:02.930 --> 00:32:06.830 is a sampling error. Now before going to the sample size calculation, 404 00:32:06.830 --> 00:32:11.930 there is a question. Diya asked where the biases 405 00:32:11.930 --> 00:32:16.930 which may arise in systematic sampling especially with periodic patterns 406 00:32:16.930 --> 00:32:21.230 in the population. So what happens is whenever there is a systematic 407 00:32:21.230 --> 00:32:25.830 sampling we are going to conduct, still it may not be a 408 00:32:25.930 --> 00:32:30.130 total representative of the population because even if we are doing it 409 00:32:30.130 --> 00:32:34.730 systematically, even if we try to remove all the biases, there is 410 00:32:34.730 --> 00:32:38.830 a chance that the samples that we have collected, they are not 411 00:32:38.830 --> 00:32:45.330 exactly the representation of the whole sample, okay? So now just 412 00:32:45.930 --> 00:32:50.130 as I have told you that we need to get the random 413 00:32:50.130 --> 00:32:55.530 numbers from for the random sampling, okay? 414 00:32:55.530 --> 00:33:00.830 So systematic random sampling we need to have 415 00:33:01.430 --> 00:33:05.830 the random numbers, okay? So I'll just share all of you. Please 416 00:33:05.830 --> 00:33:11.530 open your web browsers so that you can also do it in 417 00:33:11.530 --> 00:33:16.630 real time, okay? I hope you all can see my screen. So 418 00:33:16.630 --> 00:33:25.330 here you have to go and type open API. Okay, 419 00:33:25.430 --> 00:33:28.630 so this is a menu for open API. It is a very 420 00:33:29.230 --> 00:33:34.630 nicely prepared software. It is also available online website for various 421 00:33:35.430 --> 00:33:44.130 statistical tools. Here we have the random numbers, okay? So 422 00:33:44.330 --> 00:33:48.030 if we want to this is an example, the random number general 423 00:33:48.030 --> 00:33:53.230 they have explained this module generates a specified number 424 00:33:53.230 --> 00:33:57.630 of random integrates within a given range. To obtain decimal results, generate 425 00:33:57.630 --> 00:34:01.030 integrates and then move the decimal point as needed. We don't need 426 00:34:01.130 --> 00:34:05.530 decimal results. What we need is we need a random set of 427 00:34:05.530 --> 00:34:10.830 numbers in a particular population, okay? So the random 428 00:34:10.830 --> 00:34:16.330 number generator lowest value will be one, highest value will be how 429 00:34:16.330 --> 00:34:21.730 many cases you need, okay? So say for example, we 430 00:34:21.730 --> 00:34:27.130 need a total of 200 cases from 1,000 431 00:34:27.130 --> 00:34:30.830 population. So the lowest value will be one, highest value will be 432 00:34:30.830 --> 00:34:36.230 1,000. How many cases are required? We require 200 cases and 433 00:34:36.230 --> 00:34:39.930 format into how many columns? So five columns is okay and then 434 00:34:39.930 --> 00:34:44.630 we say calculate. So here is a list of 200 random numbers 435 00:34:44.830 --> 00:34:49.930 from one to 1,000 population. So if in the 1,000 436 00:34:49.930 --> 00:34:55.430 medical students we want to conduct a study using simple random sampling, 437 00:34:55.630 --> 00:35:00.530 we can just enumerate the number of 1,000 number of students from 438 00:35:00.530 --> 00:35:05.230 one to 1,000. We need to add the roll number and name 439 00:35:05.330 --> 00:35:09.830 the students one to 1,000 and we can use this random sample 440 00:35:09.830 --> 00:35:14.430 generator from open API and you can code this in your research, 441 00:35:14.530 --> 00:35:18.730 okay? In the methodology, you need to code that for my research, 442 00:35:18.830 --> 00:35:24.030 I used random number generator from open API. You can just 443 00:35:24.030 --> 00:35:29.030 copy-paste this link and you can just copy-paste a screenshot 444 00:35:29.030 --> 00:35:33.030 of this in your project or you or for your article publication, 445 00:35:33.030 --> 00:35:38.030 you can just mention that you used the random numbers generated 446 00:35:38.030 --> 00:35:43.330 from the open API. Okay, let's take another example of the 447 00:35:43.330 --> 00:35:48.530 diabetes study. We want to include 200 cases. The highest value 448 00:35:48.530 --> 00:35:53.930 is 2 lakh. We have 2 lakh cases of diabetes. So 449 00:35:54.030 --> 00:35:58.130 one to 2 lakh and we need to calculate 200 cases of 450 00:35:58.130 --> 00:36:03.130 diabetes, okay? So here we have if we 451 00:36:03.130 --> 00:36:07.830 have a list of 2 lakh patients of diabetes and we want 452 00:36:07.830 --> 00:36:12.830 to include 200 cases here we can just include these numbers. 453 00:36:12.830 --> 00:36:17.830 So like number 41,976, number 4,499 will be 454 00:36:17.830 --> 00:36:21.830 included in the study. So this is the random number generator and 455 00:36:21.830 --> 00:36:27.030 that we need to okay. So there is one another question. 456 00:36:27.630 --> 00:36:32.730 How to account for variability in outcomes when calculating sample sizes for 457 00:36:32.830 --> 00:36:38.130 multi-center study, okay? So this is a very technical 458 00:36:38.130 --> 00:36:43.130 statistical question. It will require a long lecture 459 00:36:43.130 --> 00:36:48.230 or even longer lecture for the variability outcome calculator when 460 00:36:48.230 --> 00:36:53.230 calculating the sample sizes for multi-center study. But in multi-center studies 461 00:36:53.230 --> 00:36:58.430 also you can use different type of sampling method 462 00:36:58.430 --> 00:37:04.330 and that can be used for the multi-center study, okay? So 463 00:37:04.330 --> 00:37:10.030 next we will come we will do a small 464 00:37:10.130 --> 00:37:15.130 sample size calculation for a 465 00:37:15.130 --> 00:37:20.430 probability sampling and for a prevalence study, okay? So 466 00:37:20.730 --> 00:37:25.630 for sample sample size calculation for descriptive study, the formula for sample 467 00:37:25.630 --> 00:37:30.030 size calculation for a descriptive study is n is equal to p 468 00:37:30.730 --> 00:37:35.830 into one minus p in the bracket z by e squared, okay? 469 00:37:35.930 --> 00:37:39.230 So what are these symbols? We will come to know about that. 470 00:37:42.330 --> 00:37:47.430 p here is the estimate of expected proportion with the variable 471 00:37:48.230 --> 00:37:53.730 of interest in the population. So p 472 00:37:54.330 --> 00:38:00.130 what we have as a p is the estimate of expected proportion 473 00:38:00.130 --> 00:38:04.930 with the variable, okay? It can be a number. Say for example 474 00:38:05.230 --> 00:38:10.230 for the study of diabetic foot ulcers, we have the cases 475 00:38:10.230 --> 00:38:14.930 of diabetes. For that, we need either the proportion of diabetes in 476 00:38:14.930 --> 00:38:19.530 India or in urban slums or in in the Indian metropolitan city 477 00:38:19.830 --> 00:38:23.330 or the proportion of diabetes in the whole world. So any of 478 00:38:23.330 --> 00:38:26.430 these proportions we can use as a p. If it is in 479 00:38:26.430 --> 00:38:31.930 the form of percentage for 10 % we will use 0.1. For 480 00:38:32.030 --> 00:38:37.130 15.4 % we will use 0.154, 481 00:38:37.230 --> 00:38:42.330 okay? q here mentioned here will be the 482 00:38:42.430 --> 00:38:46.330 one minus p, okay? So the q stands for one minus p. 483 00:38:47.130 --> 00:38:51.030 e squared is e is nothing but the precision level. We can 484 00:38:51.030 --> 00:38:56.430 take the precision of 5%, 7% or 10% more 485 00:38:56.430 --> 00:39:01.330 than 10% we should not take the e or the precision level. 486 00:39:01.330 --> 00:39:05.130 5% is the ideal level of precision that can be included for 487 00:39:05.130 --> 00:39:10.330 the sample size calculation. Four is nothing but the z squared. z 488 00:39:10.330 --> 00:39:15.530 is the statistics for desired level of confidence that is 0.05 489 00:39:15.630 --> 00:39:20.730 which comes to 1.96 for 95% confidence interval. 490 00:39:20.730 --> 00:39:25.330 z is nothing but 1.96 which is rounded off to two. So 491 00:39:25.330 --> 00:39:29.830 the z square as per this formula z square will become four. 492 00:39:30.430 --> 00:39:34.830 So this formula in simple terms it is sample size n is 493 00:39:34.830 --> 00:39:38.830 equal to four into p into one minus p which is the 494 00:39:38.830 --> 00:39:43.030 q and divided by e square, okay? So this is the formula 495 00:39:43.030 --> 00:39:48.030 that you can mention in your research methodology that sample 496 00:39:48.030 --> 00:39:52.230 size calculation you will be using the formula for sample size calculation 497 00:39:52.230 --> 00:39:55.430 from the proportion which will be n is equal to four pq 498 00:39:55.430 --> 00:40:00.230 by e square. So majority of the studies that you conduct this 499 00:40:00.330 --> 00:40:05.330 formula will be applicable. Example, if we want to study the prevalence 500 00:40:05.330 --> 00:40:10.730 of diabetic retinopathy in an urban slum, the required proportion or 501 00:40:10.730 --> 00:40:14.130 p that we will be using either the prevalence of diabetes or 502 00:40:14.130 --> 00:40:19.130 prevalence of diabetes diabetic retinopathy from a research article and 503 00:40:19.330 --> 00:40:23.430 we can code this research article in your methodology in the sample 504 00:40:23.430 --> 00:40:28.430 size calculation that a research article conducted by in this 505 00:40:28.430 --> 00:40:33.230 case a kumar at all they have reported the prevalence of diabetes 506 00:40:33.230 --> 00:40:37.530 of 10.5 % which is the p is equal to 10.5 % 507 00:40:37.930 --> 00:40:43.130 which is equal to 0.105. q will be one minus p which 508 00:40:43.130 --> 00:40:48.330 will be 0.895 and considering the absolute error 509 00:40:48.330 --> 00:40:53.530 of 5% which will be 0.05. The sample size calculation here 510 00:40:53.530 --> 00:40:57.030 will be n is equal to four pq by e square. It 511 00:40:57.030 --> 00:41:03.730 will be four into 0.105 into 0.895 512 00:41:03.830 --> 00:41:09.130 divided by e square. The e that is 0.05 or 5% 513 00:41:09.130 --> 00:41:15.030 it will come to around 151. So we will code that a 514 00:41:15.030 --> 00:41:19.830 kumar at all reported prevalence of diabetes of 10.5 and we are 515 00:41:19.930 --> 00:41:24.030 taking the sample size for this prevalence of diabetic retinopathy in an 516 00:41:24.030 --> 00:41:29.430 urban slum. Sample size that will be taken is 151 or any 517 00:41:30.330 --> 00:41:35.730 any number larger than 151, okay? So if we have 160 eligible 518 00:41:35.730 --> 00:41:39.530 cases, we will include all the 160 cases. We will just add 519 00:41:39.530 --> 00:41:44.430 one line that the sample size calculated was 151 but we got 520 00:41:44.430 --> 00:41:49.230 larger number of cases than 151 like 160. So we included all 521 00:41:49.230 --> 00:41:53.230 the 160 cases or 200 cases in our study. The sample size 522 00:41:53.530 --> 00:41:57.730 calculated from this formula it should be the minimum sample size. At 523 00:41:57.730 --> 00:42:01.830 least 151 cases of diabetes should be included in the study. Less 524 00:42:01.830 --> 00:42:05.230 than that you cannot include but more than that yes we can. 525 00:42:06.330 --> 00:42:11.730 Okay? So the take home message for from this will be 526 00:42:11.730 --> 00:42:16.730 we cannot study the whole population. So we reduce it by in 527 00:42:16.730 --> 00:42:20.630 the terms of samples. Taking a sample leads to sampling error which 528 00:42:20.630 --> 00:42:25.830 is measurable, good design and quality assurance ensure validity and appropriate sample 529 00:42:25.830 --> 00:42:30.730 size will ensure precision. Probability samples are the only one that allow 530 00:42:30.730 --> 00:42:35.330 use of statistics as we know them and for further reading you 531 00:42:35.330 --> 00:42:40.330 can go and read textbook of on research guide 532 00:42:41.730 --> 00:42:45.830 where we have topics of sample size calculation and sampling methods in 533 00:42:45.830 --> 00:42:50.830 the research. Okay? So let's just take another example for 534 00:42:51.130 --> 00:42:56.330 the sample size calculation using the open API software, okay? 535 00:42:56.330 --> 00:43:01.530 So same website open API you can just mention you can just 536 00:43:01.530 --> 00:43:05.030 take a screenshot of this. For sample size calculation, you will be 537 00:43:05.030 --> 00:43:11.030 using openapi.com for random number calculation. You will also be using openapi.com, 538 00:43:11.030 --> 00:43:16.030 okay? So I'll just the same example we will take. I'll show 539 00:43:16.030 --> 00:43:20.430 you how we came to know about the prevalence of diabetes 10.5% 540 00:43:20.630 --> 00:43:25.030 using this article by a kumar at all review article by a 541 00:43:25.030 --> 00:43:30.230 kumar at all using Google Scholar and 151 we have come from 542 00:43:30.230 --> 00:43:34.230 using this formula but using EP info it will be a little 543 00:43:34.230 --> 00:43:39.430 bit less. So we'll just go and try. So stay with me 544 00:43:39.430 --> 00:43:44.430 for few more minutes. So 545 00:43:44.430 --> 00:43:49.630 we want to do an article. We want to do 546 00:43:50.530 --> 00:43:56.330 sample size. We want to check the sample size from the 547 00:43:56.430 --> 00:44:01.930 open API, okay? So here first I search 548 00:44:01.930 --> 00:44:06.430 about prevalence of diabetes in India in Google Scholar. Okay, this is 549 00:44:06.430 --> 00:44:09.930 not normal Google. This is Google Scholar or same way you can 550 00:44:09.930 --> 00:44:14.930 search in the PubMed. So we got different articles about epidemiology of 551 00:44:14.930 --> 00:44:19.130 type two diabetes, prevalence and trends of diabetes, high prevalence of diabetes, 552 00:44:19.430 --> 00:44:23.530 epidemiology of type two diabetes in India and this is the article 553 00:44:23.730 --> 00:44:26.730 by a kumar at all where there is a prevalence of diabetes 554 00:44:26.730 --> 00:44:31.230 in India reviews of ID of diabetes atlas 10th edition. When we 555 00:44:31.230 --> 00:44:35.030 open this link by a kumar at all, we got the prevalence 556 00:44:35.030 --> 00:44:41.530 of diabetes in India and they have reported that worldwide 557 00:44:41.530 --> 00:44:47.030 prevalence diabetes affects estimated 537 million adults worldwide, 558 00:44:47.030 --> 00:44:52.130 okay? And it is around 10.5% of all adults in this age 559 00:44:52.130 --> 00:44:57.430 range. So this is the 10.5% prevalence for diabetes 560 00:44:57.430 --> 00:45:01.130 in whole world, okay? So we can use this prevalence but more 561 00:45:01.230 --> 00:45:06.830 precise will be the number of diabetes cases in India, okay? Using 562 00:45:06.830 --> 00:45:10.630 this prevalence, what we will do is we will go to the 563 00:45:10.830 --> 00:45:14.930 open API menu. We will go to the sample size. We will 564 00:45:14.930 --> 00:45:20.830 go to the proportion. In the proportion, we 565 00:45:20.830 --> 00:45:23.730 will enter new data. So this is the sample size of a 566 00:45:23.730 --> 00:45:29.330 proportion and here they have explained how this module calculates 567 00:45:29.330 --> 00:45:33.930 the sample size from the proportion, okay? Now we go to the 568 00:45:34.230 --> 00:45:39.530 enter new data. For here we need population size. It is not 569 00:45:39.530 --> 00:45:44.530 important. It is not required to change this population size. If we 570 00:45:44.530 --> 00:45:48.430 know exact population, say for example in a city of Mumbai, you 571 00:45:48.430 --> 00:45:52.130 know that Mumbai city 2 crore people are there. You can enter 572 00:45:52.130 --> 00:45:56.430 2 crore people, okay? Or just leave it as it is. Anticipate 573 00:45:56.630 --> 00:46:02.630 anticipated frequency here. We have got anticipated frequency of 10.5%. 574 00:46:02.730 --> 00:46:08.830 So here it is in the percentage, okay? 10.5%. Confidence level, limit, 575 00:46:09.330 --> 00:46:14.030 percent or absolute precision. So this is absolute precision e which is 576 00:46:14.030 --> 00:46:18.730 5%. We need to keep it that way and this design effect 577 00:46:18.730 --> 00:46:24.330 also one we had to we have to keep see. So 578 00:46:24.430 --> 00:46:30.230 for a confidence interval of 95% the sample size comes to 145. 579 00:46:30.330 --> 00:46:35.330 What we calculated was 151. It was because we have 580 00:46:35.430 --> 00:46:40.530 taken z square as four, okay? Actually z 581 00:46:40.530 --> 00:46:46.330 square is 1.96 into squared which comes to around 3.84 582 00:46:46.330 --> 00:46:51.230 some, okay? But because of that this five six number of cases 583 00:46:51.230 --> 00:46:55.830 more we have got by that formula four pq by e square. 584 00:46:55.930 --> 00:47:00.030 So this is the 95% sample size that we have got is 585 00:47:00.530 --> 00:47:05.630 145. If we want to go further to 97% confidence 586 00:47:05.630 --> 00:47:10.830 interval, sample size will be 177 or 99% confidence interval, 587 00:47:10.930 --> 00:47:16.330 sample size will be 250. So these way we can just calculate 588 00:47:16.330 --> 00:47:22.230 sample size for various research. For example, if you want to conduct 589 00:47:22.230 --> 00:47:26.630 a study for prevalence of anemia in pregnancy or any topic of 590 00:47:26.730 --> 00:47:31.830 anemia in pregnancy, we just search for research articles 591 00:47:31.930 --> 00:47:36.330 in Google Scholar. What is the prevalence of anemia in pregnancy? So 592 00:47:36.830 --> 00:47:42.430 we will just try to open if the prevalence is mentioned anywhere 593 00:47:42.830 --> 00:47:47.030 and we need to find out a number which quotes the prevalence 594 00:47:47.130 --> 00:47:51.230 that this much is the prevalence of anemia in the pregnancy. So 595 00:47:51.230 --> 00:47:55.230 here they have got overall prevalence of anemia in the pregnant woman 596 00:47:55.230 --> 00:48:00.730 of 36.8%, okay? Which is around 38%, 37%, 597 00:48:00.730 --> 00:48:06.030 sorry. So we can code this study by Muhammad 598 00:48:06.030 --> 00:48:09.530 Karami at all. So this is M Karami at all. If we 599 00:48:09.530 --> 00:48:14.030 want to cite this article, you can just click on cite and 600 00:48:14.030 --> 00:48:18.830 you can just copy the Vancouver style citations from here and you 601 00:48:18.830 --> 00:48:23.030 can just a control c or copy it and you can copy 602 00:48:23.930 --> 00:48:28.430 and paste it to your references. So M Karami at all says 603 00:48:28.430 --> 00:48:34.230 that the global prevalence of the anemia in pregnancy it is 37%. 604 00:48:34.530 --> 00:48:39.430 Now we want to calculate sample size for prevalence of anemia and 605 00:48:39.430 --> 00:48:44.030 pregnancy. We will go to the open API menu, sample size, proportion. 606 00:48:44.130 --> 00:48:48.430 In the proportion, we will go to enter new data. Anticipated frequency 607 00:48:48.430 --> 00:48:54.030 is 37%, confidence interval, confidence limit or absolute precision is 5% 608 00:48:54.330 --> 00:48:59.130 and we will just hit calculate. For 95% confidence interval, we need 609 00:48:59.130 --> 00:49:04.230 to include 359 cases of anemia, okay? 610 00:49:04.330 --> 00:49:08.330 So this is how the sample size will be calculated for various 611 00:49:08.330 --> 00:49:13.030 studies and if you have any questions or if you want me 612 00:49:13.030 --> 00:49:19.430 to guide you regarding sample size calculation, yes. So 613 00:49:21.830 --> 00:49:26.030 what is the impact of anticipated attrition rate on sample size? Yes. 614 00:49:26.030 --> 00:49:30.130 So this is a very good question by an attendee. So the 615 00:49:30.130 --> 00:49:35.630 for anticipated attrition rate, you can just include 10% 616 00:49:35.630 --> 00:49:40.730 more samples in your study. Say for example, in this particular sample 617 00:49:40.730 --> 00:49:44.830 size for this particular study, we got sample size of 359. You 618 00:49:44.830 --> 00:49:48.430 just add one line to compensate for the attrition rate or to 619 00:49:48.430 --> 00:49:52.830 compensate for the loss of data, loss of follow-up. We will include 620 00:49:52.830 --> 00:49:58.030 10% more cases. That will be 359 plus 10% will be around 621 00:49:58.030 --> 00:50:02.200 36 cases and that will be total around 400 cases. So for 622 00:50:02.200 --> 00:50:07.400 this particular study you can include 10% more cases 623 00:50:07.400 --> 00:50:11.900 and you can justify the sample size modified sample size as 400 624 00:50:11.900 --> 00:50:17.100 total number of cases. So that's all regarding different sampling techniques 625 00:50:17.100 --> 00:50:22.200 and the sample size calculation hands-on training. I again 626 00:50:22.500 --> 00:50:27.800 thank you very much assimilate by meta university and Dr. madam for 627 00:50:28.100 --> 00:50:32.700 conducting and for organizing this research. So thank you so much Dr. 628 00:50:32.700 --> 00:50:35.800 Superna ma'am and thank you so much all the participants for the 629 00:50:35.800 --> 00:50:38.100 questions and for patient listening.