mice
mice copied to clipboard
p-values imprecision for summary.mipo
Hi there, thanks for the excellent package!
I was pooling some data and noticed some (highly significant) p-values were coming out as 0 which then messes with multiple comparisons correction.
I dug in the code and spotted the following line in summary.mipo:
https://github.com/amices/mice/blob/1c1ac6fd542ae11b1ab0225699a7b1e1ea09989f/R/mipo.R#L71
I think that the 1 - pt(...)
introduces numerical problems and it's more precise to use pt(..., lower.tail = FALSE)
to get the complement.
For example:
2 * (1 - pt(abs(-11.5), df = 104))
# [1] 0
2 * (pt(abs(-11.5), df = 104, lower.tail = FALSE))
# [1] 3.059153e-20
Looks like a simple fix, I will knock up a quick PR with the code and link this issue.