Re: st: Group-mean equality test

[Nick Cox <njcoxstata@gmail.com>]

An appropriate t-test will depend on a appropriate model for the
generating process. As the data are time series, trend and various
kinds of dependence should be considered. Applying a standard t-test
as if the data were independent within 5 year blocks is at your peril.

Nick

On Wed, Aug 10, 2011 at 4:14 PM, lreine ycenna <lreine.ycenna@gmail.com> wrote:
> Hi All,
> I am looking for an easier way to calculate the mean values for each
> of my variables for different time horizons (e.g. variable x from
> 1985-1990, 1991-1996..etc) and at the same time test the signficance
> (t-stats) for the changes between the means for these periods within
> each variable itself. I manage to go from egen meanedu= mean(edu) if
> year < 1991 & year > 1979 & country==GBR. and the same for the next 5
> year period..etc. Though this only gives me the mean values for the
> time periods wanted, how do I then test the significance for the
> changes using ttest?
>
>
> year          code          edu
> 1973    GBR     0.6096253
> 1974    GBR     0.5443353
> 1975    GBR     0.5181908
> 1976    GBR     0.3762168
> 1977    GBR     0.5183734
> 1978    GBR     0.3920778
> 1979    GBR     0.4526944
> 1980    GBR     0.4694474
> 1981    GBR     0.3979259
> 1982    GBR     0.3916153
> 1983    GBR     0.2517771
> 1984    GBR     0.2545344
> 1985    GBR     0.2207009
> 1986    GBR     0.1595718
> 1987    GBR     0.30558
> 1988    GBR     0.2499627
> 1989    GBR     0.4576487
> 1990    GBR     0.5781643
> 1991    GBR     0.5855764
> 1992    GBR     0.4679879
> 1993    GBR     0.3789743
> 1994    GBR     0.6381156
> 1995    GBR     0.6055152
> 1996    GBR     0.4644643
> 1997    GBR     0.195214
> 1998    GBR     0.2751569
> 1999    GBR     0.4539562
> 2000    GBR     0.46502
>

*
*   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/