Is it 'scientifically rigorous' to study the entire population instead of selecting a sample in a study?
Saigopal is correct in that a small population that is well defined can be studied. The major advantage is that you can actually use descriptive statistics rather than predictive statistics and truly know what is going on with that population...no guess work (prediction) needed.....
By the way, these kinds of data can still be used in a kind of generalization of findings... That is, others can use those population data to generalize to other populations...but in a special kind of generalization. Rather than "extrapolation from a sample to a population" which is the way generalization works based upon probability theory, a kind of "case-by-case (or population-by population) transfer" can be used wherein results of the actual population study are used IF the decision to transfer or generalize the findings are made by the reader of the research rather than the original researcher...and IF the reader/consumer of the population study has a sufficient description of the research done and the characteritstics of the population so that she/he knows the essential operational conditions of the original population and determines whether these operational conditions also occur in the second population....This is often done for studies in Law and Medicine and is often used on cases....but could be employed in a well defined small population as well. In 1993 Firestone discussed this as has Ruddin and Kennedy....See below.
Obviously if if you've got the time and energy and resources to study the whole population you should do so because then you get the whole answer and don't need to mess around with p values or confidence intervals etc. This must be more rigorous. There is obviously no good reason to ignore data you can get! However, the real population of interest is often wider than the obvious population. For example you may have data on all instances in the past year. In practice you will usually want to extrapolate to the future so your population now includes the much more nebulous "and similar cases in the future" and of course the option of collecting all the data is now gone! The assumption behind almost all statistical procedures to extrapolate conclusions from a sample to a population is that the sample is a random one from the population, so the question you need to ask before citing p values or whatever for an analysis based on "population" data is "Can it be regarded as a random sample from the recent past and near future?" This may be a difficult question!
Always, study on whole population is better than study on sample. I you have time , energy and structure sampling frame, you should go for study of whole population
It is of course scientifically rigorous to study an entire population, but I would suggest you ask two questions.
1. Is it cost effective? This may not go down well with academics, but could you achieve the same or a similar level of accuracy through a cheaper sampling methodology.
2. Would your "entire population" survey be as accurate as you think? You don't say who you are seeking to study or how you will source your data, but no group is homogenous, and it may be that some people are more difficult to reach than others, potentially resulting in an inbuilt bias in your data.
Never assume that just because a dataset purports to be complete that it actually is. Some people assume that Census data is "entire population" data, but even that carries with it a series of confidence intervals.
One advantage of sampling is that you need to clearly address the issue of selection bias etc from the outset. These issues don't disappear with data that purports to be "entire population". If you go down the latter route, scrutinise the data collection methodology every bit as much as you would for a survey and proceed with care.
The answer is simple: yes, it is ok. It's rather the other way round as implied in your posting. The question is not if it is ok to study the entire population, but if it is ok to study a sample (given that all problems of representativity etc. enter the picture). So, if you have the possibility to study the whole population, you should do it.
What conditions justify studying the entire population instead of selecting a sample?
When the population is small & you can access / collect responses from all your respondents in that population.
I think its neccesary to do both. a sample in phase I, II and III studies. but its also neccesary to evaluate near all the population in real world evidence. this is the true and most real evidence.
If you can handle (time and money permitting) a population, then just do it, as is put forward by other colleagues, studying a population will almost always be advantageous over studying a sample of the population.
If both population and number of variables on which data is likely to be sought are small it would be worth and appropriate to consider the entire population for study. If any one of them is large chances of handling large volumes of missing data will increase and defeat the concept of studying the entire population as some observations may have to be discarded resulting in to a sample itself ultimately.
When you are the government and it is the year when your country conducts a census! (That was just tongue in cheek)
If the size of population you want to research is small and you can handle it, you can do that research...
If the size is workable, and accordingly the methodology will be different. The whole population is like a case study, no longer issues of representative sample.
To my mind you can never observed the 'population' even if you have a census as this is a theoretical concept that is not achievable in practice.
I have set out the argument and consequences of this in much more detail in a previous posting
https://www.researchgate.net/post/Are_p-values_and_significance_tests_still_meaningful_in_census_data_analysis
as I say there " the observed count should be considered to be the outcome of a stochastic process which could produce different results under the same circumstances. It is this underlying process that is of interest and the actual observed values give only an imprecise estimate of this"
In brief I think you are always dealing with a sample.
I know that a lot of people will disagree with me!
I agree with Kelvyn that basically we are dealing with samples all the time. Often, if not always (at least in my area -- the Social Sciences and Humanities) it is problematic to pinpoint precisely what/who is the entire population. It can be a philosophical question rather then a purely methodological issue.
Even if a researcher has data from the "entire population" (say, all people who learn a rare language X as a foreign language in formal settings in a country Z), we would have more learners of this language in future (say, next semester). (I do understand that we are talking about "entire population" at the time of data collection but we are likely to infer the statistical results to future "populations", so...)
An interesting question therefore could be "What conditions make a sample the entire population?".
1-population census.
2-small size of population
3- case finding studies for prevention and control for a certain disease
4- Epidemic or outbreak to know sources of this outbreak.
3 & 4 study is conducted for individual factors which need testing all persons in the population. Sample will not allow to reach your objective
The entire population is used when members of the study site are few or when its necessary to reach each member.
Its important to consider entire population to avoid sampling errors associated with sampling techniques
A census study occurs if the entire population is very small or it is reasonable to include the entire population (for other reasons)
There are several scenarios that determine the choice of population to be adopted by a researcher. A researcher has the freedom to adopt any (census or sample) depending on his/her intention and also, the population size. However, If the features/characteristics of the population are evidently seen or can be observed in the sample, the researcher can adopt sampling. On the other hand, if adopting sampling will reduce, hinder or threaten realization of the set objective or loss of necessary information, it is advisable therefore, to adopt a census study. It must be stated that care must be taken when adopting sample to avoid bias.
If you work with quantitative variables in populations, there are statistical tests that determine the sample size to ensure that statistical analyzes have consistency.
Theoretically, a researcher is at liberty to use sample or census. My experience as a data scientist, samples work if and only if they are drawn scientifically regardless of the population size. Now, I here scholars argue that census is a OK when the population is small. The catch in that statement is the word 'small'. What really is 'small population?'. Clearly, 'small' is a subjective adjective that can be interpreted relative to many parameters such as the finance ability of the researcher, geographical coverage etc. There is no standard scale to determine what is really is small
You can investigate the entire population when you have a small population size, and this is better than sampling with regard to generalization your findings. Additionally, you can avoid sampling errors associated with the different sampling techniques.
It depends upon the population. If the total population is small- then census - that is surveying all the people is ideal and it is definitely more rigorous as the population in question may not be homogeneous. If it is homogeneous and population is not small , then we do not gain any real advantage as we need to spent more time and money. One problem you come across is the response rate . If it is low then finally it becomes a sample only rather than a census. Thank you.
When the population is so small and a fraction of which may not measure what is required
It true when population is small you can use total population sampling.I faced with a similar situation where population is listed companies in Tanzania is 27.what i did i use judgemental sampling to select senior managers from those companies for questionnaire survey.Is that correct? Unit of analysis is companies.Any help or suggestions please.
If the Population is relatively small, I recommend using it as the sample.
I'm having a discussion with a Masters student who got into the same snug. The contributions here are valid. However, we find more strength in evaluating the population in question in-terms of research feasibility. In this regard, the researcher must always consider the feasibility in-terms of time and any other resources required to study the population. This will give valid reasons on whether to sample or to go for a census.
It is a good argument.In addition to that scholars are claiming if the population is less than 50 subjects, you have to go for a census or the entire population(refer to Kitchenham& Pfleeger,2002, principles of survey research-population and samples Part-5).
Sampling technique is very essential for quantitative researches. I noticed above, that most of the respected respondents recommend that you can target the whole population when you have a small size. the question here is "What do you mean by small size, " What is the minimum number that we consider it as a small sample?"
The entire population might be better than sampling, we already try to generalization sample to entire...
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology