Econometrics applied to SAS

[spraynardz90]

Hi

I'm looking to perform an event study, using econometric regressions with SAS.

Just wondering what is the difference between the following regressions:

-Logit

-Probit

-Tobit

[Ksharp]

-Logit

-Probit

They are almost the same which all do the Logistic Regression. The only thing different is Link Function is different.

Logit use logistic function, while Probit use  Normal Probability Cumulative Function. The result and explanation of both almost are similar .

-Tobit

No idea ,yet .

Good Luck.

Ksharp

[SteveDenham]

Tobit is a censored regression method, where the response variable is censored above a certain value, below a certain value, or both.  It comes in several varieties, and Heckman models are a type of tobit regression.

See: http://en.wikipedia.org/wiki/Tobit_model

Steve Denham

[Ksharp]

Hi, Steve

Really thank your explanation . You mean Tobit is used to Survival Analysis ?

I really love statistical theory, and I really hope I have enough time to study it.

But Now I have to focus on JAVA .

NO Choice.

Regards.

Ksharp

[SteveDenham]

Not really for survival analysis, although that is where you find it in SAS.

Two good examples:

Suppose you were regressing years of education (x) on income (y), but all you had for the dependent variable was incomes up to say 100,000.  This would be a good case for tobit regression.

Another is in analytical work.  Suppose the concentration of a water pollutant was determined by some method, and there was a lower limit of quantitation.  Now samples come in from various places, and you want to check the relationship between the water pollutant, and perhaps some other chemical.  The pollutant has measured values down to the limit of quantitation, but we want to develop a predictive equation using the other chemical.  Tobit regression would be one tool that has been used.

For more examples, see PROC QLIM and PROC LIFEREG.  Alternate methods that do not employ tobit analysis, but do deal with arbitrary cutpoints like these are PROC ENTROPY and PROC FMM.

Steve Denham