I am interested in evaluating differences in direct,indirect, and total
effects from the output of an SEM structural model.  How are analyses of
differences in means distinct from analyses of differences in effects?

Example 1: Communication (CM) in interorganizational exchange has been
shown to have both a direct effect on cooperation (CP) and an indirect
effect, mediated by trust (TR).  Let's say the standardized direct effect
of CM on CP is .33 (.078 standard error) and the indirect effect is .55
(.066 se).

Example 2: We now want to evaluate differences in the total effects of
communication on cooperation for two groups (e.g., vendors versus
customers).  Let's say the standardized total effects of CM on CP for
customers is .36 (se .089) and for vendors is .17 (se .036)

Questions:  Is it appropriate to use Cohen's d to evaluate the magnitude
these differences? If not, can you suggest another method?  Should
standardized or unstandardized effects be used in the analyses? Why or why
not?

[此贴子已经被作者于2005-10-31 2:32:18编辑过]