Presumably, you've thought about the issues with testing multiple hypotheses
here....
If you just wanted this once, you could do it using a simple loop within
your do file, but to make it multi-purpose, a simple program would be best.
Here's one that will do what you ask (assuming you want the p values from
simple t-tests) and put the results in a matrix. I adapted some code I
already had that did something somewhat similar -- it's pretty basic but
does include using value labels for labeling rows/columns when they are
available.
program define mymat
version 8
syntax varlist(numeric max=1) , by(str) mat(str)
confirm numeric var `by'
qui levels `by', local(bys)
local size: word count `bys'
mat `mat'=J(`size',`size',0)
forvalues i=1/`size' {
local ival: word `i' of `bys'
local ilab: label (`by') `ival'
if real("`ilab'")<. local ilab "`by'`i'"
local ilabs "`ilabs' `ilab'"
forvalues j=`=`i'+1'/`size' {
local jval: word `j' of `bys'
qui ttest `varlist' if inlist(`by',`ival',`jval'), by(`by')
mat `mat'[`i',`i']=r(mu_1)
mat `mat'[`j',`j']=r(mu_2)
mat `mat'[`i',`j']=r(p)
mat `mat'[`j',`i']=r(p)
}
}
matrix rownames `mat'= `ilabs'
matrix colnames `mat'= `ilabs'
end
you would use it for your example as:
mymat Price, by(State) mat(pricebystate)
and it would create a matrix called pricebystate with the results you seek.
(assuming that State is a numeric var with value labels --if not, encode it
first)
I'm not sure how displaying the resulting matrix with 50 columns may look,
it may be best to drop _all and use svmat to create a dataset with the
results.
Michael Blasnik
[email protected]
----- Original Message -----
From: <[email protected]>
To: <[email protected]>
Sent: Thursday, October 02, 2003 9:51 AM
Subject: st: matrix with averages and p-values
> I have a data set that looks like the following:
>
> State Price
> 1. AL 40
> 2. AL 50
> 3. AL 40
> 4. AZ 20
> 5. AZ 20
> 6. AZ 30
> 7. AZ 40
> 8. CA 20
> 9. CA 50
> 10. CA 20
> 11. CA 40
> 12. CA 50
>
> I would like to get a table with averages (on the
> diagonal), and p-value for the test that two averages
> are equal (off-diagonal):
>
>
> AL AZ CA
>
> AL avg(AL) p(AL=AZ) p(AL=CA)
>
> AZ --- avg(AZ) p(AZ=CA)
>
> CA avg(CA)
>
>
> Since I have 50 states and I have to repeat the same
> analysis for different variables and different data
> sets, I was wondering whether there is an easy way for
> doing this.
>
> Thank you,
>
> Raffaella Baldi
*
* 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/
