Message/Author |
|
|
Drs. Muthen, I was wondering if you could provide me with a little guidance on understanding a certain portion of the output derived from my SEM model completed in Mplus. My first question is what does the total r2 for the (latent) dependent variable mean (see output below). THe mIs this the total variance accounted for by the dependent variable? I wonder how the program arrived at this calculation. R-SQUARE Latent Two-Tailed Variable Estimate S.E. Est./S.E. P-Value DEP 0.493 0.010 49.409 0.000 Also, I would like to get an estimate for all the other predictors in my model, pwb, ewb, and swb, not just the total for the entire model. Would it be sound methodology to just square the STDYX Standardization estimate for each of these indicators to get the r2 coefficient? Example: Estimate S.E. Est./S.E. P-Value DEP ON MHC_EWB -0.431 0.023 -18.607 0.000 Would the r2 just be -0.432 which comes out to 0.1849? Thanks. |
|
|
R-square is given for each dependent variable. It is not possible to get the part of R-square attributable to each covariate. The are not independent. |
|
|
I would like the R-square value for each of my latent variables with respect to the outcome variable. I am having trouble determining where in the output these values are. For example, R-square is provided for the indicators in my model and only 3 of 6 of the latent variables in my model. E.g., R-Square Observed Variable Est... Y1... F1... I would like to know the variance accounted for in my outcome variable by each of my latent variables separately. Could you assist? |
|
|
An overall R-square is given for each dependent variable in the model. It is not broken down by covariate. |
|
|
There is no option to determine the R-square for independent variables? |
|
|
R-square is the variance explained in a dependent variable by a set of independent variables. |
|
|
Thank you, I know what R-square is. I would like to know how much of the variance in my outcome variable is explained by EACH of my latent variables: some of these latent variables are dependent. That is, even though they are hypothesized to predict the outcome, there are other variables in the model that also predict those variables. However, there are 2 independent variables in my model, and I would like to know how much of the variance in the outcome variable is accounted for by those 2 IVs. The M+ output has not provided these values. It is only indicating how much variance is accounted for in the outcome by the other DVs. This does not make logical sense to me. Could you please help me. |
|
|
Never mind, thank you for your help. It seems that you misunderstood my original question. I will take my output to stat support on my campus for assistance. Regards |
|
Xu, Man posted on Thursday, March 29, 2012 - 5:57 am
|
|
|
I'd like to have R-square statistic for all dependent variables in the model I run. It is already clear that I am not able to get standardised output. In this case, is there still a way to get R-square please? thanks! |
|
|
R-square comes as part of the STANDARDIZED option. If this is not available, R-square is not available. |
|
Xu, Man posted on Thursday, March 29, 2012 - 7:10 am
|
|
|
Thank you. Would it be sensible or feasible to hand calculate with the model constraints? |
|
|
If the reason you don't get STANDARDIZED is that you have random slope in your model, R-square is not defined because the variance of y varies as a function of x. If the reason you don't get STANDARDIZED is that you are using XWITH, see the following FAQ on the website where the computation of R-square is discussed: The variance of a dependent variable as a function of latent variables that have an interaction is discussed in Mooijaart and Satorra |
|
Back to top |