Anonymous posted on Friday, April 01, 2005 - 2:35 pm
Quick question, I hope. Anyway to bootstrap the standardized factor loadings? The unstandardized are great, but we really report the standardized more often. So, I was wondering if there is any way to perform this type of test. Thanks
in brief my task: I've a moderated mediation model with observ. var. only which has only n=46 and therefor low power. I wanted to use a 90% bcbootstrap. Can you suggest another idea? Second question: is the model result also generated by the bootstrap sample e.g. estimator for path b or only the confidence intervalls and the estimator in that section?
We will add 90% bootstrap in the future but for now, I have no suggestions. Only standard errors and confidence intervals are bootstrapped not parameter estimates.
marie posted on Wednesday, October 02, 2013 - 4:32 pm
My reviewer asked me to use a bootstrapping to test my indirect effects in addition to the DELTA method. My sample size is small (140). I have missing cases so I have been using MLR in my path analysis. Bootstrapping is not available with MLR. I am using Mplus 6.2. I was wondring if it was available in the later versions or there is some other reason why the bootstrapping won't work with MLR. Is it acceptable to delete the missing values and then use ML with the bias-corrected bootstrap? The data does not seem to violate the normality assumption.
You can only use BOOTSTRAP wit ML and not MLR etc. because BOOTSTRAP affects the standard errors not the parameter estimates. The parameter estimates for ML, MLR, etc. are exactly the same. It is only the standard errors that differ. So use ML.
marie posted on Saturday, October 12, 2013 - 6:29 am
Thanks. Using the commands for bootstrapping in the users' guide, I got everything I wanted except for the the Standard errors of the standardized path coefficients (direct effects). All I have is a column entitled "StdYX Estimate" with the standardized path coefficients. I am asked about the effect size. I can thus provide the standardized path coefficients but no SE from the bootstrapped results. How can I ask Mplus to give me the SE for those?
Hello, I have a few questions I was hoping for help with.
1. Is there any way to obtain bootstrapped confidence intervals (either bootstrap or bcboostrap) with clustered data (using TYPE=Complex)?
I have no replicate weights. I tried requesting REPSE=Bootstrap to generate the replicate weights followed by BOOTSTRAP=1000, as on p. 458 of manual, but I receive the same error message that "The BOOTSTRAP option with weights is only allowed with TYPE=COMPLEX when replicate weights are present or when REPSE=BOOTSTRAP is requested."
2. When bootstrapping is requested in ANALYSIS and CInterval, with nothing following CInterval in parentheses, in OUTPUT, are the confidence intervals bootstrapped? My understanding is that the default for CInterval is symmetric rather than bootstrapped CIs. However, p. 38 of the manual indicates that "When both the CINTERVALS and BOOTSTRAP options are used, bootstrapped confidence intervals are computed. (p. 38).
I seemed to get slightly different results when using CInterval versus CInterval(bootstrap) in conjunction with bootstrap command in ANALYSIS.
Thank you so much Dr. Muthen for your quick reply and for the clarification. One follow up question- can REPSE=BOOTSTRAP be used to generate replicate weights that could then be used to obtain bootstrapped SEs and CIs with clustered data? Or is this not appropriate? I confess I am not that familiar with weights in this context.
I am using WLSMV and I also have a categorical covariate. I wanted to estimate the CI of the indirect effects and I got the following message:
BOOTSTRAP and BCBOOTSTRAP confidence intervals are only available for the regular results of the model and MODEL INDIRECT. No confidence intervals are printed for the standardized results of the model and MODEL INDIRECT.
I was wondering why it does not provide CI for the standardized results? I searched the manual but I couldn't find the answer. Is it because of the covariate?