Message/Author |
|
|
Hi, I have the following Model, using the WLMSV estimator. Categorical are all. C_pi_bin is the IV (binary) C_K by vn167Ar vn167Br vn167Cr; C_S by c_iss1 c_iss2 c_iss3; c_wahl on c_pi_bin; C_K on c_pi_bin; C_S on c_pi_bin; c_wahl on C_S; c_wahl on C_K; C_K on C_S; MODEL INDIRECT: c_wahl ind c_pi_bin; c_wahl ind C_S; When estimating this model, I only get the estimates in the STDYX, but no s.e. and p-values. Why is this so? When I rewrite it, so that c_pi_bin is a latent variable with only 1 indicator, that looks like this: C_PID by c_pi_bin@1; c_pi_bin@0; I get all the results, but the warning that my IV is treated as continious, but is binary. Can I interpret the results nevertheless, because it is an IV? Thanks for your help. |
|
|
I would not do C_PID by c_pi_bin@1; c_pi_bin@0; because you add the assumption of underlying normality for your binary covariate. When you have a binary covariate you don't want to use STDYX for the slopes on this covariate, but STDY. |
|
|
Hello, I have the same problem as Martin Schultze. When I add a certain construct with 3 indicators to my model and add more than one path from this construct the p-values and s.e. would disappear in the output. I am a newbee in Mplus and may overlook something obvious but could you tell me what is the reason for that? Thanks for your help! |
|
|
With WLSMV, standard errors and p-values for standardized parameters are not available for conditional models. |
|
|
Thanks Dr. Muthen, but what does "conditional" mean in this context. I am wondering why just this construct would make a difference while all others won't. Could you give me a hint where I can find this issue in the user manual? Thanks for your help! |
|
|
A conditional model is a model with covariates. |
|
|
Thanks for the answer! Sorry to ask again but what does this mean? I tried to find out with the help of the Users Guide, Google etc. but could not find a hint. Does it mean: 1. Something is wrong with my dataset? 2. Some constellation which makes the model not computable? 3. Do I have to leave the construct out of my model in order to get p-values and fit indices or can I make this work somehow? |
|
|
Please send input, data, output and license number to support so we can see exactly your situation. |
|
RuoShui posted on Wednesday, April 02, 2014 - 4:55 pm
|
|
|
Dear Dr. Muthen, You said in the above posting "With WLSMV, standard errors and p-values for standardized parameters are not available for conditional models". Can I be reasonably confident in applying the p-values for unstandardized parameters to the standardized parameters? I am hoping to use the standardized parameters to compare the relative predictive value of each covaraite. Thank you very much! |
|
|
Their p-values usually agree, but not always. |
|
|
Hi, I have the same issue. No p-values or std errors for sdyx & std estimates. How can I know the p values? Is it safe to assume that p-values for unstandardized parameters will be the same for standardized parameters Thank you very much! |
|
|
The p-values are different for unstandardized and standardized parameters because their sampling distributions are different. If you don't get p-values, you may be using an older version of Mplus where they were not available. |
|
|
Dear Linda, Thank you for your response. I am using Mplus 7 and my model is; Emo by mams1x mams2x mams3x mams4x mams5x mams6x mams7x mams8x mams9x mams10x; Beh by mams11x mams12x mams13x mams14xr mams15x mams16x; Emo Beh on Ethnic Language SEN CLA FSM IDACIcat; Emo with Beh; Is there something I can do to get the S.E.s & p-values? |
|
|
If they are available, they are printed. |
|
Back to top |