Using effect size or stdYX?
Message/Author
 victor china posted on Thursday, April 21, 2016 - 12:27 am
i am running the following model: (it has both categorical and continuouse variables) i wish to learn how each variable (total effect) affects "fin", and to be able to determine which variable is more important relative to the others)
USEVARIABLES lnTl lnTW fin ini tac pos lnTc LavgTw;
CATEGORICAL = fin ini tac pos;

ANALYSIS: ESTIMATOR = WLSMV;
parameterization = theta;
MODEL:
lnTc ON lnTw@0 lnTl ini tac@0 pos LavgTw@0;
ini ON lnTw@0 lnTl LavgTw;
tac ON lnTw@0 lnTl ini LavgTw@0;
pos ON lnTl tac;
fin ON lnTw@0 lnTl ini tac pos lnTc LavgTw@0;

MODEL INDIRECT:
fin ind pos;
fin ind ini;
fin ind lnTl;
fin ind lnTw;
fin ind lavgtw;
lnTc ind ini;
lnTc ind lnTl;
lntc ind lavgtw;

OUTPUT: tech1 STANDARDIZED SAMPSTAT ;

victor
 victor china posted on Thursday, April 21, 2016 - 12:28 am
i would like to add that:
all my categorical variables are 0 and 1.
the output produces only stdYX estimates, is it correct to present the stdYX values?
i only wish to be able to determine which total effect from each relevant variable is more important relative to the other ( in respect to "fin").
can i use effect size or stdYX ?
 victor china posted on Thursday, April 21, 2016 - 12:57 am
i know you wrote:
With binary covariates, StdYX should be adjusted so that the standardization uses only the standard deviation of y, not the standard deviation of x.

but when a path consists of both categorical and binary variables, how do i interpret the total effect? with which estimate?
 victor china posted on Sunday, May 01, 2016 - 1:59 am
any ideas?