0

I am currently analysing a series of data collected annually.

each year varies, however, it is convenience sampled.

one year, for example, has a sample of 2427 positive cases, of a known population of 251800.

I am using binary logistic regression to investigate the cases by subgroup and demographic.

I have been asked by my supervisor to provide a post-hoc power analysis and don't understand which method to use, given I know the population, and the cases.

PowerLess
  • 1

1 Answers1

2

I would politely approach your supervisor with some resources which argue (correctly) why post hoc power is meaningless. See my answer here to a similar question.

In short, post hoc power is just a transformation of the p-value. You get no additional information from doing a post-hoc power calculation.

Demetri Pananos
  • 36,121
  • thank you for your fast response.

    I have previously stated that post-hoc power analysis is not appropriate in a known environment, however, my supervisor countered with the fact that i have uncovered a data set that shows the previously unknown total population.

    I have since given an analysis based on standard error, however, the suggestion from my supervisory remains and unfortunately it may be a hurdle to overcome.

     – PowerLess Jan 27 '20 at 17:12