Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: Request help in interpretation of p values for multiple linear regression OR bug in program
From
Maarten Buis <[email protected]>
To
[email protected]
Subject
Re: st: Request help in interpretation of p values for multiple linear regression OR bug in program
Date
Mon, 9 Sep 2013 09:16:34 +0200
With computer programs you get exactly what you ask for, which is not
necessarily the same as what you want. It is therefore essential that
you _exactly_ tell use what you typed into Stata. The command you
typed does not correspond with the output you showed. I am guessing
that you forgot to include the factor variable notation, but we should
not be guessing.
What you showed us is certainly no bug in Stata; it just shows
different tests for different null hypotheses. It should not come as a
surprise that if you ask Stata different questions you'll get
different answers. I think the way forward for you is to (for now)
forget about calling variables "significant", and instead explicitly
formulate your hypotheses. The term significant can be a useful
shortcut, but as so often with shortcuts they often end up costing
more time then doing it right from the start. So your regression
output looks at each comparison of a level with the baseline level
separately, while -testparm- looks at all of them together. Neither is
right or wrong, they just represent different questions. It is up to
you to decide which quesition you want to answer.
-- Maarten
On Sat, Sep 7, 2013 at 10:54 PM, Michael Stewart
<[email protected]> wrote:
> Hi all,
>
> I am running a multiple linear regression equation for a normally
> distributed variable (y) with var1 & var2 as independant indicator
> variables.
>
> I am using stata version 12.
>
> I am trying to figure out the disconnect between the displayed p
> values(in regression table from stata output) and results of the
> testparm command
>
> Truncated output and results of the testparm command are shown below :
>
> reg y var1 var2 var3
> -------------------------------------------------------------------------------
> | Linearized
> Y | Coef. Std. Err. t P>|t| [95% Conf. Interval]
> --------------+----------------------------------------------------------------
> var1 |
> 2 | .0903065 .1012825 0.89 0.373 -.1082807 .2888937
> 3 | .5836366 .1637454 3.56 0.000 .262577 .9046963
> 6 | -.1474897 .153524 -0.96 0.337 -.448508 .1535287
> |
> var2 |
> 2 | .0597458 .1689991 0.35 0.724 -.271615 .3911066
> 3 | .1010293 .1568688 0.64 0.520 -.2065473 .4086059
>
>
> testparm i.var1
>
> ( 1) 2.var1 = 0
> ( 2) 3.var1 = 0
> ( 3) 6.var1 = 0
>
> F( 3, 3117) = 4.42
> Prob > F = 0.0042
>
> VAR1: It is significant per testparm command but only second level
> (var1 level2) is significant per the stata output.
>
> So , how do I interpret then output: Do I say var1 is significant OR
> only level2 of var1 is significant ??
>
> OR Is there a problem in stata progam ?? (unlikely )
>
> Can someone help me out please.
>
> Thank you in advance for your time and effort
>
> --
> Thank you ,
> Yours Sincerely,
> Mike.
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/faqs/resources/statalist-faq/
> * http://www.ats.ucla.edu/stat/stata/
--
---------------------------------
Maarten L. Buis
WZB
Reichpietschufer 50
10785 Berlin
Germany
http://www.maartenbuis.nl
---------------------------------
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/