Re: st: RE: RE: time trend in a panel with very SHORT time dimension

[Steven Samuels <sjhsamuels@earthlink.net>]

Jun, why do you think five years is " too short?  Standard books on  
longitudinal data (e.g. Diggle, Liang,Zeger, 1996) show series as  
short. Even series of two or three periods can be analyzed.  Perhaps  
you are thinking of time series data. If you wanted to time series  
analyses (-help time series-), you would indeed need many more  
periods, but nobody is suggesting a TS approach.

Your analyses and  Alan's suggestion will indeed test for a trend,.  
If you want to model the trend, I suggest a full longitudinal  
analysis.  Plot the trajectories to examine their form. How do you  
know the trend is linear?  How do you know that, as your models  
imply, the slopes are identical?

If the five-year intervals start in different years, you might have a  
problem generalizing to a trend over the whole period.  However you  
can add a "year" of start as a predictor in a regression model to  
account for that.

-Steve

P Diggle, KY Liang, S Zeger 1996. Analysis of Longitudinal Data.  
Oxford Press.


On Feb 13, 2009, at 11:30 AM, Jun Zhou wrote:

> Hello Al,
> Thanks a lot for your reply.
> My biggest concern here is the time dimension is too short, which  
> might affect the inference on the time trend. Can -somersd- address  
> this issue?
> Thanks again.
> Jun
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner- 
> statalist@hsphsun2.harvard.edu] On Behalf Of Feiveson, Alan H. (JSC- 
> SK311)
> Sent: February 13, 2009 9:17 AM
> To: statalist@hsphsun2.harvard.edu
> Subject: st: RE: time trend in a panel with very SHORT time dimension
>
> You could try a non-parametric within-cluster comparison using - 
> somersd- . This avoids distributional assumptions.
>
>  somersd expense year,funtype(wcluster) cluster(FirmID) wstrata 
> (FirmID)
>
>
> Al Feiveson
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner- 
> statalist@hsphsun2.harvard.edu] On Behalf Of Jun Zhou
> Sent: Thursday, February 12, 2009 7:31 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: st: time trend in a panel with very SHORT time dimension
>
> Dear all,
>
> I have an unbalanced panel dataset with 5 years observations from  
> 2000 firms.
> I would like to test whether there is a positive time-trend in one  
> variable (e.g. expense).
> I tried the following two ways:
>        xi: areg expense year, absorb(FirmID) cluster(FirmID)
>        reg expense year, cluster(FirmID) Both methods show that the  
> slope (i.e. the coefficient of 'year') is positive and  
> statistically significant. Can I conclude that there is a positive  
> time trend in variable 'expense'?
>
> I understand that it is improper to try to test time trend in a  
> time series of 5 observations, but I am not sure whether 2000 firms  
> can allow me to test the time trend although each firms have no  
> more than 5 observations.
>
> Thanks a lot for your help.
> Jun
>
> *
> *   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/
>
> *
> *   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/
>
> *
> *   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/

*
*   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/