

Path model with poisson outcome 

Message/Author 

Erin Casey posted on Friday, May 26, 2006  11:59 am



I am estimating a path model with a poisson distributed (count) outcome, 5 mediators (1 categorical, 4 continuous) and 2 continuous IVs. The data is from a complex, stratified survey. I have three questions: 1) Do I interpret the StdYX for the paths to the count outcome identically to the StdYX for paths from the IVs to the mediators? 2) I am not getting fit statistics (CFI, RMSEA). Is it possible to get these in the context of a poisson regression? 3) Am I correct in specifying the binary mediator as such in the syntax (given that its both an IV and a DV?) Thanks much for your help!! Syntax: STRATIFICATION is region; WEIGHT is gswgt3_2; CLUSTER is psuscid; COUNT is perpw3; categorical is perpw2; ANALYSIS: Type = complex missing ; integration = montecarlo; MODEL: perpw3 on perpw2 h3ma4r impsex meandel expviol sexdrink; impsex on h3ma3r h3ma4r; meandel on h3ma3r h3ma4r; expviol on h3ma3r h3ma4r; sexdrink on h3ma3r h3ma4r ; perpw2 on h3ma3r h3ma4r; h3ma3r; h3mar4; OUTPUT: sampstat standardized patterns; 


1. Only the metric of the dependent variable is different. 2. No. These test statistics are available only when the means, variance, and covariances are sufficient statistics for model estimation. 3. Yes. Note that when it is a dependent variable it is treated as binary, but when it is an independent variable is is treated as continuous. 


I am estimating a path analysis with a zero inflated Poisson dependent variable using MLR. I have 3 questions similar to those above: 1) I'm not receiving model fit statistics (e.g., chisquare, RMSEA, CLI/TLI, etc.). Are these not available for zero inflated poisson DVs? 2)When I attempt to evaluate indirect effects, the model will not run and I receive the following error message: "MODEL INDIRECT is not available for analysis with ALGORITHM=INTEGRATION." I have not included "ALGORITHM=INTEGRATION" in my input. Am I correct in thinking that indirect effects can not be examined with MLR? 3)Finally, when I request modification indices, I receive the following error message: "MODINDICES option is not available for ALGORITHM=INTEGRATION. Request for MODINDICES is ignored." Again, I do not have "ALGORITHM=INTEGRATION" in my input. Any idea why I can not receive modification indices? Thank you 


1. Chisquare and related fit statistics are available for only when means, variances, and covariances are sufficient statistics for model estimation. 2. You can use MODEL CONSTRAINT when MODEL INDIRECT is not available. 3. Numerical integration is required for your model. It is not a choice. 

Back to top 

