|
![](http://www.statmodel.com/discussion/icons/csc/1px.gif) |
Constraints in Logistic model |
![Previous](http://www.statmodel.com/discussion/icons/csc/navgen_arrow_left.gif) ![](http://www.statmodel.com/discussion/icons/csc/1px.gif)
|
Message/Author |
|
Mark Shevlin posted on Wednesday, October 26, 2005 - 9:34 am
|
|
|
I wish to impose an equality constraint on the slopes for five predictor variables (and leave one unconstrained) in a univariate logistic regression model. I have tried the following syntax without success - the model runs but the estimates are not equal. When I remove the constraints the df's and LL are the same. Hope you can help, Thanks TITLE: DATA: FILE IS "psychosis.dat"; VARIABLE: NAMES ARE deplt1 naplt raped phy_att phy_ab neglect molested; USEVARIABLES ARE deplt1 naplt raped phy_att phy_ab neglect molested; MISSING ARE ALL (99); CATEGORICAL ARE naplt ; ANALYSIS: TYPE = logistic ; ESTIMATOR IS MLM; ITERATIONS = 1000; CONVERGENCE = 0.00005; OUTPUT: CINTERVAL; model: naplt on raped phy_att phy_ab neglect molested (1); naplt on deplt1 ; |
|
|
Equality constraints are not available for TYPE=LOGISTIC; You can obtain logistic regression with equality constraints by removing TYPE=LOGISTIC; and ESTIMATOR=MLM; from the ANALYSIS command and adding ESTIMATOR = MLR; |
|
|
Many thanks Linda, that works perfectly. |
|
Back to top |
|
|