Prior to my study I had determined a small sample size for my research analysis. As a result of my dataset and research questions, a larger sample size was necessary to run my analysis. My urr is stating that using a large sample size is not permitted and that I should justify. I have done so using other research articles and even had statisticians say that is weird of him to ask me to justify. Please any thoughts and directions is appreciated
Hi Naomi,
that was my first idea: You have planned your sample assuming a relevant effect size, a minimum power and a fixed alpha niveau. If you work with a much greater sample, then even smaller effects become significant, that may be absolutely negligible for your research question. So after raising your sample size, you have to calculate the effect size of each significant effect to prove his significance with regards to content.
Best
Brian
Hi Naomi, It is not uncommon that sample size needs to be "justified". Research can cost a lot of money or consume a lot of resources in general. Consequently, many research and funding bodies want you to calculate the apriori sample size that is needed to estimated your hypothesized effect. This can be achieved by means of power analysis. Sample size calculations for surveys are often handled somewhat differently in practice. Applications and examples can be found here:
https://onlinecourses.science.psu.edu/stat414/node/245.
A nice free tool for conducting power analysis is G*Power and can be downloaded here:
http://www.gpower.hhu.de/
.
maybe the question is why you state
(a) that prior analysis determined a need for a small sample size
(b) that it came out that a larger sample size was necessary
in other words, what went wrong in step (a) ?
.
now, if what went wrong in step (a) is understood and corrected, i see no reason not to rerun the analysis in step (a) and it should give a larger sample !
.
Hello Brian, Thomas and Fabrice, I greatly appreciate your answers. I will use the power analysis as suggested by Dr Volken. Can I send my paper for you all to see what I actually did for my analysis? may be that will help. Thank you again.
Dear Ms Azunie,
Based on my knowledge, learnings and wide experience, I believe you may have done things the wrong way round.
Allow me to explain.
In carrying out big data analysis ( or data science research / machine learning projects ) there is a massive ongoing methodology debate among the top experts in the field. Explaining the whys and wheretofores of this ongoing methodological debate would fill a 6500 page encyclopaedia, so complex and complicated is its rationale and underpinnings. However, it is a debate ( well, series of debates) that I have been following avidly for the past 14 years.
Understanding these series of debates will allow you to understand and resolve the question that you have posted on here.
Do let me know if you would like me to explain further.
Yours sincerely.
Dalton Coker
Can you tell us more about what kind of data analysis you are doing and what you mean by a "small" and a "large" sample?
By the way, "big data analysis" is just a fancy term for " very complicated & time-consuming statistical work"...
There is wide variation in approach and methodology to carryout a pilot study and real study data collection and analysis
Hi Naomi.
Increase the sample can be justified by the central limit theorem.
A small study is often done to establish relevant parameters and help to revise and develop the instruments that are then used in the larger study. The pre-study actually serves to make the larger study more relevant.
- Badges
- Science topic