|From||Richard Williams <Richard.A.Williams.5@ND.edu>|
|Subject||Re: st: Parameters returned by -lincom- after -logistic-|
|Date||Mon, 09 May 2005 07:30:21 -0500|
At 12:34 PM 5/9/2005 +0100, you wrote:
Hello again (especially StataCorp)p. 37 of the Stata 8 svy manual gives returned values and some formulas. e(df_m) and e(df_r) don't quite correspond to the d.f. reported by the F stat but I suspect there is some trivial formula that shows the relationship. I know that, in writing my own program, I had to make sure e(df_r) was available when svy was used, otherwise I got z's instead of t's reported.
One more query re using -lincom- after -svy- commands in Stata 8. How are the degrees of freedom calculated? I cannot find anything about this on Page 21 of -[SVY]-. I ask because a lot of the examples in -[SVY] lincom for svy- have t-statistics instead of z-statistics, and, as I understand it, the programs -svymean-, -svyratio- and -svytotal- use a method dependent on whether -e(complete)- is -complete- or -available-, with a vector degrees of freedom in the latter case and a scalar degrees of freedom in the former case.