----- Original Message -----
From: <[email protected]>
To: <[email protected]>
Sent: Thursday, June 26, 2003 10:33 AM
Subject: st: Re: t-tests for differences in means using pweights
> I don't think my original posting was clear enough. What I have are two
unique
> samples, one from the first wave of a survey and the other from the next
wave.
> In each wave respondents were asked the same question, and I'm looking at
> aggregate means on this question for each wave of the survey. I'm trying
to
> determine whether the difference in the mean value of the variable among
> respondents in the first wave is significantly different from the mean
among
> respondents in the second wave (after weighting the responses with
pweights).
> The problem is that the samples are unique, so I can't use the "svymean"
> or "svylc" commands, but I also can't use the "ttest" command b/c it does
not
> accept pweights. I hope this is clearer than the first post below, and
again I
> would appreciate any thoughts. Thanks in advance.
>
> Pat Sharkey
> Doctoral student in Sociology and Social Policy
> Harvard University
How about using a dummy variable and svyreg. Using the auto dataset,
suppose sample 1 is the price of car domestic cars and sample 2 is the price
of foreign cars and the sampling weight is the weight variable.
. use "C:\Stata8\auto.dta", clear
(1978 Automobile Data)
. svyset [w=weight]
(sampling weights assumed)
pweight is weight
. svyreg price for
Survey linear regression
pweight: weight Number of obs =
74
Strata: <one> Number of strata =
1
PSU: <observations> Number of PSUs =
74
Population size =
223440
F( 1, 73) =
0.13
Prob > F = 0.72
13
R-squared =
0.0015
----------------------------------------------------------------------------
--
price | Coef. Std. Err. t P>|t| [95% Conf.
Interval]
-------------+--------------------------------------------------------------
--
foreign | 298.4742 833.3626 0.36 0.721 -1362.415
1959.363
_cons | 6500.577 503.7304 12.90 0.000 5496.644
7504.51
----------------------------------------------------------------------------
--
The t-stat on the dummy variable tests if the mean price of foreign cars is
equal to the mean price of domestic cars.
So, the null hypothesis of Ho: mean(Domestic) - mean(Foreign) = 0 can be
rejected at usual significance levels.
Scott
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
