Op 01/24/2005 03:14 PM schreef Salkov, Rick:
I have a simple dataset consisting of 2 variables a and b. I suspect that the fit is not linear so I used fraqpoly.
The Stata help was a little vague on the generated formula.
Here is the run.
. fracpoly regress b a
........
-> gen double Ia__1 = X^-2-1.194 if e(sample)
-> gen double Ia__2 = X^3-.7661 if e(sample)
(where: X = a/100)
Source | SS df MS Number of obs = 46
-------------+------------------------------ F( 2, 43) = 60.96
Model | .011742256 2 .005871128 Prob > F = 0.0000
Residual | .004141086 43 .000096304 R-squared = 0.7393
-------------+------------------------------ Adj R-squared = 0.7272
Total | .015883342 45 .000352963 Root MSE = .00981
------------------------------------------------------------------------------
b | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
Ia__1 | -.0001748 .0004779 -0.37 0.716 -.0011385 .0007889
Ia__2 | -.0138219 .0014766 -9.36 0.000 -.0167998 -.0108439
_cons | .1115939 .0018996 58.75 0.000 .107763 .1154247
------------------------------------------------------------------------------
Deviance:-297.9678. Best powers of a among 44 models fit: -2 3.
After gen x = a/100, I thought that the regression formula would be:
b = -.0001748*a^(x^-2-1.194) - .0138219*a^(x^3-.7661) + .1115939
omit a^ (twice) and you should be OK.
Arnold
*
* 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/
