topic Re: sas proc calis in Statistical Procedures

sas proc calis

thanoon — Thu, 27 Feb 2014 06:02:29 GMT

i want to estimate parameters of sem by using weighted least square method but i have some question through the following sas program:

1- i want to write my own weight matrix how??

2- how can i write initial values?

3- how can i change link function here?

proc calis nobs=1000 data=MVO method=WLS;

var X1-X10;

lineqs

X1 = a1*F1+e1,

X2 = F1+e2,

X3 = F1+e3,

X4 = F1+e4,

X5 = a5*F2+e5,

X6 = F2+e6,

X7 = a7*F3+e7,

X8 = F3+e8,

X9 = a9*F4+e9,

X10 = F4+e10,

F2 = F1+d1,

F3 = F1+d2;

std

e1-e10 = The1-The10,

d1-d2 = psi1-psi2;

cov

e3 e5 = The11,

e4 e6 = The12;

weight;

run;

thanks in advance

Re: sas proc calis

SteveDenham — Thu, 27 Feb 2014 14:42:26 GMT

The documentation for CALIS lays these out fairly clearly (and I would NEVER say that I am an expert in PROC CALIS).

1. In the PROC CALIS statement, use INWGT= to input the weight matrix.

2. In the PARAMETERS statement, initial values can be specified in parenthesis after the parameter name, for example: a1 (0.2), if you wanted an initial value of 0.2 for parameter a1.

3. There is no link function in PROC CALIS.

Steve Denham

Re: sas proc calis

thanoon — Fri, 28 Feb 2014 14:29:27 GMT

thanks alot mr. denham

i want to use categorical data with SEM then i must use link function in SEM like logit function and probit function i dont find these types of link function in sas but in other program available how can i solve this problem??

many thanks

Re: sas proc calis

SteveDenham — Fri, 28 Feb 2014 18:13:50 GMT

Transform the data prior to analysis. Unfortunately, if you are doing the multivariate distribution with a cumulative logit, that can get rather difficult. However, there are factor analysis parts to CALIS that I know nothing about, but which might accept categorical data.

Steve Denham