topic Re: power and sample size calculation in Statistical Procedures

power and sample size calculation

aw016 — Sun, 23 Apr 2023 04:02:42 GMT

I want to double-check my proposal sample size and power calculation methods

- aim 1, I have 335 participants, cross-sectional study, linear regression, the estimated effect size is 0.45, single-side test, which ways to calculate achieved power? My method is using t-test to calculate.

- aim 2, retrospective cohort study, the estimated effect size is 0.45, single-sided test, which ways will be appropriate to calculate achieved power?

Thank you!

Re: power and sample size calculation

SteveDenham — Thu, 27 Apr 2023 12:56:18 GMT

See the PROC POWER documentation for code to do these.

SteveDenham

Re: power and sample size calculation

aw016 — Thu, 27 Apr 2023 15:19:40 GMT

Hi Steve, thank you for your reply! I have PROC POWER code. However, I guess my question is more related to logic for calculating instead of calculation tools. For example, if the independent variables and dependent variables are continuous. Would it be appropriate if I calculate the achieved power in this way: categorizing independent variables into tertiles --> estimated effect size is 0.45 between 3rd tertile and 1st tertile for dependent variable --> get power around 90% based on sample size = 335?

Re: power and sample size calculation

SteveDenham — Thu, 27 Apr 2023 17:49:24 GMT

Well, you did the analysis, so you should have everything you need to calculate a retrospective power (and don't get me started on why that is a bad idea.). I would avoid categorizing a continuous variable for this. This is actually a.ONECORR analysis, since regression is just fancy correlation. You mention an effect size of 0.45. If that is an f-squared estimate, what happens when you plug into this code:

proc power;
   onecorr dist=fisherz
      nullcorr = 0
      corr = 0.557186 /* this is the square root of 0.45/(1+0.45), in other words, the square root of R-squared obtained by rearranging f-squared */ 
      ntotal = 335
      power = .;
      sides = 1
run;

If your effect size is obtained some other way (omega squared, for instance), then ONEWAYMEANS is probably better suited to the task.

SteveDenham

Re: power and sample size calculation

aw016 — Thu, 27 Apr 2023 20:18:23 GMT

Thank you so much, Steve! I was asked to show the achieved power by professor lol, although 335 is all I have.

Re: power and sample size calculation

SteveDenham — Mon, 01 May 2023 18:34:42 GMT

Given the various effect sizes I plugged in, 28 is going to give at least 80% power for a one-sided test, at an effect size of 0.45 (R squared, or.0.55 F squared) and alpha=0.05.

SteveDenham