generate abil2 = abil*abil
regress educ motheduc fatheduc abil abil2
* Here we use the "lincom" command to test a linear combination of coefficiencts
* In particular, we are interested in whether an additional year of mother's
* education has the same effect as an additional year of father's education.
* There is some suggestion that the effects are not the same.  The p-value of
* the t-ratio is p = 0.053 which is barely larger than 0.05.  There is a 
* statistical difference at the 10% level.  Maybe the mother's education is
* of slightly greater influence than the father's education.    
lincom motheduc - fatheduc
* Here is alternative way to test the same hypothesis using the F-statistic.
* We get the same p-value.
test motheduc - fatheduc == 0
* Here we add two additional explanatory variables to our equation: tuit17 and 
* tuit18.  Let's examine their joint significance.
regress educ motheduc fatheduc abil abil2 tuit17 tuit18
* Both of the tuit variables are singularly insignificant.  However, let us
* test the joint significance of the two variables.  
test tuit17 tuit18
* Noting the collinearity between the two tuition variables.
corr tuit17 tuit18 
generate avgtuit = (tuit17 + tuit18)/2
* Even taking the average of the tuition variables does not provide a 
* very satisfactory result.  Tuition continues to have an unexpected positive
* coefficient.   
regress educ motheduc fatheduc abil abil2 avgtuit