Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: RE: extract t-values


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: RE: extract t-values
Date   Fri, 7 Nov 2008 13:47:23 -0000

Quite apart from several other comments, some repeated, note that quite
what these P-values mean collectively is highly unclear. Overlapping
windows of time series cannot be independent of each other! You would be
fighting with at least three issues simultaneously, overlap; inference
for related time series; and multiple comparisons! Conversely, if you
just want the P-value as an indication of strength of relationship, then
go for a measure of strength of relationship that is less fragile and
more directly related to the (social) science of what you are doing. 

Nick 
n.j.cox@durham.ac.uk 

Bastian Steingros

 
> The thing is I run many regressions. My dataset consists of data from
> 1950 to 1995. I want to run regression  for 5-year-time-windows only,
> that is, the first time window is 1950-1955 etc. 
> To put all the statistics  from the regression table into my dataset
> simplifies my task because I then can reduce my sample to those time
> windows where all the regression coefficient are at least significant
> at the 5% level (for example by using: bys window: keep if
> p_value=<0.05).
> 
> That is the reason why I want to calculate the t-values and, in the
> next step, the p-values separately. 

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index