topic Link Functions for Logistic Regression SAS in SAS Programming

Link Functions for Logistic Regression SAS

aranganayagi — Tue, 05 Nov 2019 04:20:16 GMT

hi,

could you please help to understand where should i use three link functions - logit, probit and Complementary log-go.

I want to know in which situation i should use each one of these. Thanks for your help in advance

Re: Link Functions for Logistic Regression SAS

Ksharp — Tue, 05 Nov 2019 11:49:23 GMT

proc logistic data=sashelp.class;
model sex=age weight / link=probit;
run;

Those link function are similar .

Complementary log is for the very large data like : 999999 ( I guess ,Better check Documentation).

Re: Link Functions for Logistic Regression SAS

aranganayagi — Tue, 05 Nov 2019 13:55:52 GMT

I checked few documentation, still it is not clear. I want to know at which place I should use different link functions

Re: Link Functions for Logistic Regression SAS

Reeza — Tue, 05 Nov 2019 16:31:11 GMT

Link functions are not a SAS topic, you'd need to consult a textbook on logistic regression or generalized linear models.

Basically, if the relationship between the predictors and dependent variable is not assumed to be linear you need to use a different relationship. So you pick the link function that best models your data.

To summarize thebasic ideas, the generalized linear model differs from the general linear model (of which, for example, multiple regressionis a special case) in two major respects: First, the distribution of the dependent or response variable can be (explicitly) non-normal, and does not have to be continuous, i.e., it can bebinomial,multinomial, orordinal multinomial(i.e., contain information on ranks only); second, the dependent variable values are predicted from a linear combination of predictor variables, which are "connected" to the dependent variable via a link function. The general linear model for a single dependent variable can be considered a special case of the generalized linear model: In the general linear model the dependent variable values are expected to follow the normal distribution, and the link function is a simple identity function (i.e., the linear combination of values for the predictor variables is not transformed).

Source: http://www.statsoft.com/Textbook/Generalized-Linear-Models

Exactly how to evaluate and decide which is appropriate is likely a full course so beyond the scope of a post IMO.