RE: st: Differencing when dependent variable only defined every 4th year

[Nick Cox <n.j.cox@durham.ac.uk>]

Sure, but he still has only one result per election, so I don't see where within-election standard errors come from. 

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

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Richard Herron
Sent: 24 November 2011 19:20
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: Differencing when dependent variable only defined every 4th year

He only wants to carry-forward or fill-in certain independent
variables, correct?

@Nick -- Do you think that clustering standard errors by election
cycle would reduce the problem from the correlation in the
carried-forward ind vars?

On Wed, Nov 23, 2011 at 14:50, Nick Cox <n.j.cox@durham.ac.uk> wrote:
> This would, I think, repeat every election year four times. The coefficients might be (about) right, but the df and P-value would be quite wrong. Of course they are probably wrong any way unless the model captures the data generation process, but the sample size is surely just # of election years.
>
> In fact that's a worry independently. How many parameters are being fitted here to data for 1960(4)2008 i.e. just 13 distinct values of the response!
>
> Nick
> n.j.cox@durham.ac.uk
>
> Herron
>
> Would you like to carry forward election year data for the following
> three years? The following should replace the empty data with the
> previous election year's data.
>
> tsset state year
> replace x = l.x if (x == .)
>
> On Wed, Nov 23, 2011 at 13:13, Sebastian Barfort <sb3730@nyu.edu> wrote:
>
>> I've encountered a problem in a time series regression I want to do. I have a panel with 49 US states running from 1960-2008. All my independent variables have observations for all years. However, the dependent variable is defined only every other 4th year (it's presidential election data). The regression I want to run is the following:
>>
>> y(t)-y(t-4)=alpha+beta1(x1(t)-x1(t-1))+beta2(x1(t-1)-x1(t-2))+beta3(x2(t)-x2(t-1))+beta4(x2(t-1)-x2(t-2)) etc etc...
>>
>> But how can I do this in Stata? I've defined the data as time series with t=year the time variable, but can't seem to find the right way to go around the problem.

*
*   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/