during review of published article it was found that most of the paper do not mentioned how sample size was calculated. in our institute it is very mandatory to explain the sample size calculation with formula to get ethical clearance. are we acting differently than the other institutions.
In general it is not relevant at the point where a manuscript is published. At publication it is too late to do anything. All you are doing is showing that before the study you thought about the problem of sample size. In reading your manuscript I simply will not care about this issue. I might be interested in knowing if your collected data can prove that sufficient sample was collected independent of whether or not your treatment effect was significant.
Despite rampant use of sample size calculators we still have articles like this one: http://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.0020124
I have problems in using a sample size calculator as an ethical tool. You are saying that if my calculator tells me that I need a sample size of 673 that if I achieve a sample size of 673 that my findings will be correct? True? Valuable? What have you done to prove that the values that you entered into the sample size calculation were valid in your particular case. They are an educated guess at best, simply wrong at worst.
Of is not relevant at all.. the als last of large numbers tells any number above say 30-40 is enough.. the image is clear then and wil nog change anymore.. compare of to tossing a dice 10 times, then probabilities still mag change, but whether you toss it 30 or 1000 times.. the image remains. And even when testing s hypothesis around 40 is optimal as when sample side increases anything wil je significant, nonsense even too
Dear Laman Khanal,
I also almost never use sample size calculators or formulas because I find it to be irreevant in most of the research. It might be different in different areas of research but in my area it is much more important to mind the validity and reliability of your measurement instrument than the sample size. Validity and reliability makes sure that you measure what you wanted to measure and that you measure it the right way. The sample just has to be big enough for all the variables to work. I use the pilot testing for that and when I have enough sample units that my model works, I stop the data collection. The exeption is when you are triying to descriptively establish the representation of a certain phonomenon or a state of the matter in a certain time or area. But when the purpose of the research is to determine the relations between variables, the sample size is just not that important. It is not irrelevant (I always report it), but the size can be determined by common sense rather than calculators or formulas. Also it is a matter of cost. Sample is also determined by cost of data collection and it is quite acceptable research limitation.
I hope my thoughts help you at least a little in resolving this issue you are thinking about.
Thank you for your question that got me thinking about that and gave me an idea for a research.
Best regards and all the best!
Running an experiment until you get a significant result is one form of p hacking.
http://psychology.wikia.com/wiki/P-hacking
http://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.1002106
@Ragie, You can ask this question as one of your own, but it is not appropriate to ask this sort of question as an "answer" to someone else's question. I wish you the best of luck, but please do not do this again.
Many researchers do not include that information because they anticipate it might not be asked from them by potential reviewers. And the reason why reviewers themselves do not ask for that information is that they do not feel it is of high relevance in contrast to other issues (e.g. data collection procedure, common method bias, validity and reliability of the measurement instruments, etc.).
Dear Laxman Khanal,
This is a very interesting question, and as I could see from the several answers above, we tend to answer it according to our own experience, once there is no a formal protocol that can be used to all journals.
It is clear that maybe the main concern should be related to the validity and reliability of what is intended to be measured, as well as if there is the aim of having a study considered representative of a certain phenomenon. However, the published articles are not published only for senior researchers, neither for those already have expertise in a certain field.
Once, I had a professor, who used to say that most of the studies published in the management/business field have some kind of problem with sampling. According to him, people have concern with constructs, dimensions, variables, measures, methods of analysis, and everything else, and forget the basic work of doing sample.
I have being reading several articles published in different journals, mainly in the Marketing and Management field, which do not present any information about how the sample was definied, and do not present justifications for the choices related to it as well. Another thing that call my attention is that most of the articles do not mention the cities or regions where the studies were carried out. They simply inform that the study A or B were realized at country X or Y, about the consumption behavior K or W, the product/service S or T, and doing that is like those studies were representative of that country, behavior or product/service.
Well, maybe we should write some more lines about sampling in the published articles. Who knows, the readers will appreciate that.
All the best!
Francisco Vieira
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation