1
$\begingroup$

I am doing a study looking at the effects of a certain virus in rats. We are looking at certain concentrations of bio markers in the blood. I came into this project part way through and no power analysis was done by the project lead before hand. We originally were going to arbitrarily use 10 rats in both the vehicle and viral group. However, some of those rats died leaving only 8 in each group. I performed a power analysis on the effects we found using those 8 rats and found that we would need about 19 rats (11 more) to achieve a power of .8. Of course at the moment there are no significant differences. I know ideally we would run a new study with the 19 rats per group. However, this will not be possible due to financial limitations. Would it be possible to run 11 more rats and add their data to the 8 rats we have already done. Nothing significant has changed since we first performed the experiment and we would be careful to do the experiment the same way. Would it be ethical to do this?

$\endgroup$
3
  • $\begingroup$ What's a "vehicle group" ? $\endgroup$ – Rodolphe Feb 27 '20 at 10:32
  • $\begingroup$ Basically a control group. $\endgroup$ – Ryan Feb 28 '20 at 2:30
  • $\begingroup$ There is nothing (statistically) unethical about an experiment that is fully and honestly documented and reported. The concern about redesigning the experiment part-way through is discussed at stats.stackexchange.com/questions/310119. $\endgroup$ – whuber Jan 31 at 0:42
1
$\begingroup$

If your results with 8 rats had been statistically significant, what would you have done? Would you have done a power analysis and determined you didn't have enough power and then run 11 more rats in each group? I suspect not.

If you do this you are capitalizing on chance, and increasing your type I error rate. Your p-value no longer means what you think it means, if you do this.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.