I like the chapter in _Applied Linear Regression Models_ by Neter,
Kutner, Nachtsheim, and Wasserman on understanding and calculating
"extra sums of squares". This is the basis for a difference in
r-squared test. I think looking at -help nestreg- will be useful.
Best,
Alan
2008/11/12 SM <[email protected]>:
> Hi ,
>
> I have to compare the following (nested/ hierarchical?) models in
> order to test whether the variables included in second model
> significantly increase the goodness of fit.
>
> 1) Y = α0 + α1 BV + α2NI + ε
> 2) Y = ß0 + ß1 BV + ß2 NI + ß3 TANG + ε
>
> My questions are the following:
>
> I. It's not clear to me the difference between nested and hierarchical
> regressions. Is defined hierarchical a regression in which I simply
> add new variables ? If it so, my models (1-2) should be hierarchical.
> Conversely, is nested model a regression in which I decompose the
> components of a variable? For example, if I add a third model, in
> which TANG is decomposed in TANG1 TANG2 TANG3.
>
> 3) Y = g0 + g1 BV + g2 NI + g3 TANG1+ g4 TANG2+ g5TANG3+ ε
>
> II. In order to test whether the goodness of fit of 2 model
> significantly increases, I employ the following test, found in
> econometric textbook (considering model 2 = full model):
>
> Δ R2/ Δdf = F (Δ R2 , df full)
> (1- R2full)/df full
>
> Is it correct ?? Can you give me some specific reference on this test ?
>
> III. In Cox's book "A gentle introduction to Stata" page 216, test of
> semipartial correlation of variables is performed by pcorr2. Even if
> the author suggests to install it with findit pcorr2, I am not able to
> find the program anywhere. Can anybody suggest how can I install it
> ???
>
> Thanks so much.
>
> Kind Regards,
>
> Serena
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/