Hi Thomas,

I just saw your helpful answer at https://stats.stackexchange...

I am currently in need of a reference on how exactly the "large enough n" in the asymptotic equipartition property scales with the error parameter. I would be grateful if you could direct me to one, and / or provide a short explanation if possible. Thank you for your time and help!

Hi Thomas,

I try to understand why, when I perform a svyttest (for 2 independent samples) , my df is not equal to degf(design)-2. For example, if we run

data(api)
dclus2<-svydesign(id=~dnum+snum, fpc=~fpc1+fpc2, data=apiclus2)
(tt<-svyttest(enroll~comp.imp, dclus2))

we have df = 36, but if we run

degf(dclus2)-2

we obtain 37.

It is stated on the p. 129 of the Reference manual that "Degrees of freedom are degf(design)-1 for the one-sample test and degf(design)-2 for the two-sample case."

Thank you,
Mariia Samoilenko

Because there are missing values in 'enroll'

Also: it's better to email questions -- there's a Maintainer email field in the package DESCRIPTION for a reason

Thanks Maria and Thomas for the insightful comments :)

https://stats.stackexchange... Hi Thomas. Looking at your great answer here. I am doing something similar. I am running 1000 Monte Carlo simulations. My question is from my own code - https://github.com/winash12... Calculating the confidence intervals for the 900th, 950th and 999th simulation - is that equal to saying greater the larger the simulation size those particular simulation trials are equivalent to the 90th , 95th , 99th significance levels ?

That sounds right