-1

I'm attempting to do a power analysis on occurrence data (0.0 -> 1.0; i.e. the number a sites a species was present out of the total number of sites sampled for a single season), where I want to see what power I have to detect a change of 20% (two-sided test) from the most recent season's worth of data to the next (a future sample). For example, "...Dry season 2023 has a proportion of 0.6, what power do I have to detect a 20% change (up or down) from 0.6 for next year?" The data comes from the same 47 sites each season (paired data) and I'm using the package 'pwrss'. My questions is:

If I'm hoping to do a two-sided test, what would I put for p0?

Since I think this is a one-sample test, according to the documentation, "p0" = constant to be compared (a proportion), but is that 0.2 (20%) or 0.6 + or - 20%; (=0.72, 0.48)?

What does this syntax mean?

library(pwrss)

pwrss.z.prop(p = 0.6, p0 = 0.48,
             alpha = 0.05, power = NULL,
             n = 47,
             alternative = "not equal",
             arcsin.trans = TRUE)
Nate
  • 411
  • 2
  • 10
  • 2
    I didn’t downvote, but could see why someone (with less of a tolerance threshold) did for a few reasons - you may find this link useful: [Why is “can someone help me?” Not an actual question?](https://meta.stackoverflow.com/questions/284236/why-is-can-someone-help-me-not-an-actual-question). Good luck! – jpsmith Aug 21 '23 at 18:21
  • 1
    p0 is the alternative proportion. It is the probability you would specified in the hypothesis testing. Remember to select alternative="equivalent" and margin =0.12 in this example. – one Aug 21 '23 at 18:44
  • In my case, I think p and p0 are the upper and lower bounds of my chosen change in occurrence data: +20% = (0.6+(0.6*0.2))=0.72 and -20% = (0.6-(0.6*0.2))=0.48. This is how I interpret "expected" and "constant to be compared (a proportion)" and these are interchangeable as p and p0 since it's a two-sided test. – Nate Aug 21 '23 at 23:33

0 Answers0