1
$\begingroup$

I'm a novice in research and wondering if I need to calculate power to determine my sample size.

My study is about collecting data from all patients that presented to a clinic for a period of 6 months. Since I am collecting data from all the presentations, would this still require power analysis? If so, could someone guide me on how to do a post hoc analysis as I heard this is the most suitable for a retrospective audit. I believe this is done after the data has been collected to check if the sample size is effective. So would my research proposal just include that a psot-hoc analysis will be completed after data collection has been completed..?

Happy to hear any suggestions if another power analysis option is more suitable.

Improve this question
$\endgroup$
1
  • $\begingroup$ Why do you ask about post-hoc power when you haven't started the experiment? (I guess this from the phrase "my research proposal".) There may be opportunity to do a-priori power analysis instead. Why have you picked 6 months to collect data? Since longer study means more patients presented at the clinic, you can design (the length of) the study so that it achieves the desired power. $\endgroup$ Commented Oct 21, 2023 at 20:43

1 Answer 1

Reset to default
5
$\begingroup$

Post hoc power analysis is generally not recommended. See e.g. this article which says that they are "completely determined by p-value'' and demonstrates this. If the p value is high, power must be low and if the p value is low, power must be high.

More generally, the purpose of power analysis is to tell you how many subjects you need in your sample to have a good chance of getting a significant result if the null is, in fact, false by a certain amount.

Also see Hoenig & Heisey (2001), The Abuse of Power: The Pervasive Fallacy of Power Calculations for Data Analysis, in The American Statistician, 55, 19--24 [DOI], and more references cited there.

Improve this answer
$\endgroup$

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.