1
$\begingroup$

Anyone know why when you run a SEM model in SAS using proc calis (or tcalis) you do not get p-values for the parameter estimates? It does supply a t-value however.

Two popular SEM packages in R, 'sem' and 'lavaan', both give p-values for the estimates but they use a Z test statistic.

$\endgroup$
2
  • 1
    $\begingroup$ I am not an expert on SAS or SEM, but if you get t-values, can't you calculate p-values yourself? $\endgroup$ – mpiktas Jan 11 '11 at 4:24
  • $\begingroup$ Yes which is what I do, but in every other SAS procedure the p-value is given along with the test statistic. $\endgroup$ – Glen Jan 11 '11 at 14:46
2
$\begingroup$

@mpiktas is right and knowing the value of the test statistic ($t$ or $z$) allows you to know which parameter estimate is significant at the desired $\alpha$ level. In practice, the $t$-statistic is equivalent to a $z$-score for large samples (which is often the case in SEM), and the significance thresholds are 1.96 and 2.58 for the .05 and .01 $\alpha$ levels. Most of the time, $p$-values are interesting when comparing models; as shown in this nice tutorial on Structural equation modeling using SAS, by Y H Chan, giving $t$- or $z$-statistic with associated critical values at 5% should be enough, IMO.

$\endgroup$
1
  • $\begingroup$ Thanks. I should have made it clear that I know how to calculate a p-value given a test statistic but I find it strange that they don't output the p-value for you as in almost all the other SAS procedures. $\endgroup$ – Glen Jan 11 '11 at 14:47
1
$\begingroup$

If you're still following this: the te option (Proc Calis te) outputs pvalues.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.