Programming the statistical procedures from SAS

Non linear regression on binary data (0,1)

Accepted Solution Solved
Reply
New Contributor
Posts: 2
Accepted Solution

Non linear regression on binary data (0,1)

Hi guys, this is my first post on this forum. I need some advice. I have done a number of analyses on binary data (taking on the value 0,1) where a logistic model is appropriate. The output is a sigmoid function for the probability of being 0 or 1 that ranges from a lower asymtote of 0 to an upper asymtote of 1. I have a problem now where my data is binary (0=male, 1=female) and the relationship is sigmoid, but with an expected lower asymtote of 0.5 and an upper asymtote of 1. What procedure exists in SAS to take the binary data as input to estimate the probability of being male or female. The predictor variable is continuous -- temperature. The more general question is, I guess, what procedure in SAS allows you to estimate the parameters of a any specified function p (of being male or female) as a general non-linear regression against a continuous predictor variable. Sort of like PROC NLIN but on binary data with a binomial error structure.

Any suggestions?


Accepted Solutions
Solution
‎11-25-2014 04:55 PM
Respected Advisor
Posts: 4,606

Re: Non linear regression on binary data (0,1)

Try using proc probit. It is similar to proc logistic but allows the specification of a 'natural response rate' with option C=0.5.

PG

PG

View solution in original post


All Replies
Solution
‎11-25-2014 04:55 PM
Respected Advisor
Posts: 4,606

Re: Non linear regression on binary data (0,1)

Try using proc probit. It is similar to proc logistic but allows the specification of a 'natural response rate' with option C=0.5.

PG

PG
New Contributor
Posts: 2

Re: Non linear regression on binary data (0,1)

That worked a treat, thanks. Not sure why I did not look there first?

Thank you so much.

ods graphics on;

proc probit data=combined c=0.5 optc plots=(predpplot ippplot);

model sex = temp / dist=logistic;

output out=results p=p_hat;

run;

ods graphics off;

Analysis of Maximum Likelihood Parameter Estimates

                       Standard   95% Confidence     Chi-

Parameter  DF Estimate    Error       Limits       Square Pr > ChiSq

Intercept   1 -80.3676  24.0010 -127.409 -33.3266   11.21     0.0008

TEMP        1   2.3631   0.7007   0.9898   3.7365   11.37     0.0007

_C_         1   0.5047   0.0191   0.4673   0.5421

☑ This topic is SOLVED.

Need further help from the community? Please ask a new question.

Discussion stats
  • 2 replies
  • 239 views
  • 0 likes
  • 2 in conversation