Jeff Wardell posted on Wednesday, December 31, 2008 - 5:24 am
I am conducting a path analysis using MODEL INDIRECT to test for specific indirect effects. I requested bias corrected confidence intervals because some variables in the model are skewed. However, the confidence intervals for many of the indirect effects do not contain the parameter estimates for the indirect effects. Often, the parameter estimates are larger than the upper limit for the confidence interval. This seems odd to me. I was wondering if this may be a function of the skewness of the variables, such that the estimates are so biased as to be outside the bias-corrected confidence bands. If you could please shed some light onto this issue I would really appreciate it.
Thank you for your reply Dr. Muthen. I am using an older version (v 3.1). Does this mean the values given for the confidence interval limits are not correct? Do you think this will effect my pattern of results for the indirect paths (paths that are actually non-significant may have CIs that do not contain zero due to this printing problem?) Thank you for your time.