Using Stata for Survey Data Analysis

ovtest [, rhs] 
This test is the same as estimating y = xb+zt+u and then testing t=0. If the rhs option is not specified, 
powers of the fitted values are used for z . Otherwis, the powers of the indiependent variables are 
used.. Examples of the test are:
regress y x1 x2 x3 
ovtest 
tests significance of powers of predicted y 
ovtest, rhs
tests significance of powers of x1, x2, and x3 

---

Minot
 
 
 
 
 
Using Stata for Survey Analysis 
Page 52 
 

hettest  
Another assumption behind regression analysis is that the variance of the error term is constant across 
the sample. When this assumption is violated, the problem is called heteroskedasticity. This 
command tests for heteroskedasticity.
hettest [varlist] 
This command tests t=0 in Var(e)=s^2exp(zt). If varlist is not specified, the fitted
values are used for z. If varlist is specified, the variables specified are used for z. 
This test is also known as the Breusch-Pagan test for heteroskedasticity. Examples are:
regress y x1 x2 x3
hettest 
test whether variance related to predicted y 
hettest x3
test whether variance related to x3 
Example 24 gives the result of some tests related to the regression analysis shown earlier. The 
testparm command tests the hypothesis that both the region coefficients are equal to zero (that region 
does not influence rice expenditure). The hypothesis is rejected, meaning that the regional 
coefficients are jointly significant. The ovtest  rejects the hypothesis that there are no omitted 
variables, indicating that we need to improve the specification (prices would be a good start). And 
finally, hettest indicates that there is heteroskedasticity which needs to be addressed.