Today we are going to talk about some aspects of sampling methods see sampling is really required whenever you aredealing with a very large population andyou want some quick information now lookat the definition of sampling thesampling is a procedure by which somemembers of the population are selectedand they are supposed to be therepresentative of the entire populationsee if you have a population of peoplelike this and you are looking at aportion of them and that's called asample I would like to rather introduceyou to some concepts and one of them isthe study population what do you mean bystudy population the study population isthe population to which the results ofthe study are to be inferred say forexample how many injections do peoplereceive each year in India the studypopulation in this case is the entirepopulation of India suppose yourresearch question is how many needlesticks health care workers experiencedhere in India then the study populationbecomes healthcare workers of Indiasuppose if your study question is howmany hospitals have my needle sticksprevention policy in India then yourstudy population in this case becomesall hospitals of India this sample whichwe select should be representative ofthe population for which we require ananswer and this representation should bein accordance to the seasonality the dayof the week the time of the way whetherit's urban or its rural are it shouldrather match the composition of age sexand other demographic characteristics ofthe population see now let's ratherintroduce you to someconcepts or terminologies that are oftenused in the sampling parlance what do wemean by sampling unit sometimes it'scalled basic sampling unit we assumethey saw the elementary unit that willbe sampled that could be people orhealthcare workers or hospitals as wehad seen in a example what do we mean bysampling frame the sampling frame is alist of all sampling units in thepopulation and what do we mean bysampling scheme the sampling scheme is amethod used to select sampling unitsfrom the sampling frame so now there aredifferent ways how of why we shouldrather do the sampling populations if wehave in a resources you can probablystudy the entire population but stilleven if you have the resources it's notwise to study the entire populationbecause often the population is verylarge and a large population when youare going to rather collect informationone of the major constraint could be thetime you may require a lot of time tocollect the information and you may sayI will employ a lot of people to do thatbut what would rather probably happen isif you have lots of people collectinginformation that could be lot of interimserver variations which could rather addon to a tremendous amount of error andunfortunately you cannot measure theamount of such errors so it often eitherhappens that by doing a sample surveyyou often get accurate information theinformation that you get from samplesurveys are more accurate than theinformation you do on a large-scalepopulation studies so a population couldbe rather than an entire universewhereas a sample could beselected and small regions now let'slook at a practical example suppose theMinistry of Health of a country X wantsto estimate the proportion of childrenin elementary schools who have beenimmunized against childhood infectiousdiseases contest rather imagine you knowthe proportion of children of allelementary schools who have beenimmunized against childhood infectionsof a country so that's a task but one ofthe day conditions that he has put isthat task must be completed in one monthso the objective is to estimate theproportion of immunize children and youwant the results in a month's time nowlet's take a look at the different wayshow you can rather get this informationor in other words what are the differenttypes of samples that could either goodsee broadly speaking the sample could bea non probability sample or aprobability sample what do you mean by anon probability sample this probabilitysample is the property of being selectedthat is a sample the probability ofbeing selected for your study is notknown it could be a convenient sample ofpurposive sample you just ratherconvenient whatever the region that isconvenient to you close by to your placeyou can rather go and the see firsthundred people that you come across thatcould be a convenient sample what couldhappen that sample would be biased or itcan rather give either a best or a worsescenario people rather know it's aconvenient location you may get relatedtheir results very different from yourlocation which is not very convenient orwhich is very remote and difficult toapproach and also some of these seethese are all very subjective samplesand you to derive some objectivecriteria from a subjective samples it'salways rather difficult okay and butnevertheless these samplingnonprobability sampling methods stillrun useful and that's our beingsensibly used mainly to generatehypotheses are to prepare for moresystematic probability samples now let'srather look at what do you mean byprobability samples in a probabilitysample every unit in the population hasa known probability of being selectedwhat's the advantage this is the onlysampling method that allows to drawvalid conclusions about the populationit removes the possibility of bias inselection of subjects and also ensuresthat each subject has a knownprobability of being chosen it allowsapplication of statistical theorybecause many of the statistical teststhat you do it insist on a randomsampling and these tests are valid onlyif the samples are a random sample okayI would like to do that introduce you toa concept called sampling error nosample is a perfect mirror image of thepopulation always you know you when yout pick a sample from a population andwhen you look at the results it may notbe exactly the same as the results inthe population but fortunately themagnitude of error could be measured interms of probability in the case ofprobability samples so this is expressedby standard error of mean or proportionor differences and that is a function ofthe sample size and then the variabilityin the measurement so sampling error isa very important component in samplingtheory which helps us in identifying thesample size and things so so on nowlet's look at some of the popularsampling methodologies that are employedin sample surveyslet's leather look at the first simplerandom sampling which is as the namesuggests it's a very simple samplingprocedure very easy to understand inwhich every individual sampling unitshave got a an equal chance of beingincluded into a sample how do you dothat we number all the units and werandomly draw units so the advantagethis is has I mentioned it's very simpleand sampling error is also very easilymeasured major limitation of this is youneed to have a complete list of allunits many times it may not be availableand also sometimes you may get a samplewhich is very different from the wholepopulation may not be veryrepresentative of the population see anexample of a simple random samplingcould be if you have the list of all sayabout this what hit names you pick arandom numbers of 9 18 32 and 40 sothese are all the names that areselected as your samplenow the next sampling type is asystematic sampling systematic samplingwhat is rather than it is the initialsampling unit is picked by random andthen every Cait unit from that from yourpopulation are examined so a unit isdrawn at every K units and every equalchance of being selected for each of theunit so you calculate the samplinginterval called K which is divided by an divided by the number of sample setsthat you require and you draw a randomnumber which is less than or equal to Kfor starting and draw every K units fromthe first unit what are the advantage itensures representativity across the listit's easy to implement you give a workerthatsay you start from this house and every10,000 you go and rather see and coverall the houses it's very easily it'sbeing done if there is some sort of acycle of some specific characters thatyou are studying then you might probablyget a sample which is very typical insystematic sampling and also some of thestatistical measures that you are goingto rather compute it's difficult whenyou are going to really have systematicsampling where you do not try thathaving exact formulas you may have torather use them approximate formulas theexample of a systematic sampling is thisis you see in the first the red house isselected and then every 8th house fromthat is rather selected and this iswhere all the red houses in these housesare your selected samples there is asampling method called stratifiedsampling the principle of rate assessesyou classify population into homogeneoussubgroups which are called strata andyou draw a sample from in each stratacombine the results of all this data toget an idea of the whole population sothe advantage of it Isis is it's moreprecise if variable associated withstrata and all subgroups represented arerepresented allowing for separateconclusions about each one of themsuppose you know your natural stratacould be male and female so you have aan estimate for male and you have anestimate for a female and you can havean estimate for a combined male andfemale for the whole population but thedisadvantage is a sampling error isdifficult to measure and that could be aloss of precision if you are windra thathave a lot of strata and for each stratayou have small numbers in it example ofa stratified sampling is is suppose ifyou want to rather estimate thevaccination coverage in your country onesample drawn from each region not eastsouth and west and the estimatecalculated for each of the stratum andthen at the end you can rather wait thestratum according to the theestimated the size of the regions okayso another important type of samplingwhich is very popularly used and withyou in the health surveys in research iscalled cluster sampling the principle ofcursor sampling is that a random sampleof groups or a cluster of units and allour proportion of units are included inthe selected clusters say an advantageis it's simple we don't require a listof units and less of travel or resourcesare required because you are going torather collect a cluster and you aregoing to rather see only within theclusters and the disadvantages is if theclusters are homogeneous then you mayhave rather have may result in a largedesign effect the all the people in thesample may have very homogeneous resultswhich could rather result in designeffect and sampling error is difficultto measure in a cluster sampling thesampling units is not a subject but agroup or a cluster of subjects theassumptions here is this thatvariability among the cluster is minimalthe variability within each cluster iswhat is observed in the generalpopulation so now how these clustersampling is Genesis's usually it is doneas a two stage approach in the firststage a probability proportional to sizethat is select the number of clusters tobe included computer cumulative list ofall the population in each unit with agrand total divide the grand total bythe number of clusters and obtain thesampling interval choose a random numberand identify the first cluster and thesampling interval and identify thesecond cluster and so on and byrepeating the same procedure identifyall the clusters okay once your clustersare identified then in the second stagein each cluster you select a sanamrandom sample using a sim sampling framebecause as I had rather mentioned youearlier on simple random sampling whenyou want to rather do a simple randomsampling you need to have all the listsof their sampling frame so in a smallcluster it is possible for you toformulate the sampling frame and you canreselect people from that sampling frameon a random basisanother important sampling methodologythat's Rovner employed is called amulti-stage sampling and in this multistation especially in a very largesuppose you want some estimates for atthe national level you need to either dosampling in several chain samples andseveral statistical units are there theadvantages is there is no completelisting of the population is requiredand it is most feasible approach forlarge proportions populations ok thedisadvantages is there are severalsampling units and sampling error attimes it's very difficult unless youfollow certain very specificmethodologies for selecting at eachstage some of the key issues that Iwould like to rather bring to your thisis we cannot study the whole populationso we sample it whole populationstudying it ASIS could rather implyresult in inaccurate results so thetaking sample leads to sampling errorbut which is easy which is easilymeasurable and we don't have a measurefor non sampling error whereas we have ameasure for sampling error good designand Quality Assurance ensure validityand appropriate sample size will ensureprecision so the probability sample orthe only one that allow the use ofstatistics as we know them and so it'salways advantage to use a probabilitysample so that you canthey'll have a valid conclusion aprocess conclusion and also you canemploy statistical tests on them thankyou so much