Message/Author |
|
Lydia Zhang posted on Monday, February 05, 2018 - 5:37 pm
|
|
|
Hi, I am trying to use the manual 3-step approach to examine the relationship between latent classes and children's developmental outcomes over time. In the third step, I conducted growth-curve modeling to see if child trajectories differ by latent classes. Outcome variables are at three time waves (OUTCOME1-OUTCOME3). Latent variables are U1-U4. The first step works well but there are errors in the third step. Could anyone help me to check why it is not working? Thank you! |
|
Lydia Zhang posted on Monday, February 05, 2018 - 5:41 pm
|
|
|
------------------------------------- TITLE:INPUT FOR FIRST STEP; DATA: File is test_1.dat ; Names are OUTCOME1 OUTCOME2 OUTCOME3 U1 U2 U3 U4; Categorical = U1 U2 U3 U4; Classes = N(3); Usevar=U1 U2 U3 U4 ; Auxiliary=OUTCOME1 OUTCOME2 OUTCOME3 Analysis: Type = Mixture; Savedata: File = test_PROB.dat; Save = cprob; |
|
Lydia Zhang posted on Monday, February 05, 2018 - 5:42 pm
|
|
|
TITLE: INPUT FOR THIRD STEP DATA: File is test_PROB.dat; Names are U1 U2 U3 U4 OUTCOME1 OUTCOME2 OUTCOME3 cprob1 cprob2 cprob3 cprob4 N; Usevar=OUTCOME1-OUTCOME3 N; Classes = C(3); nominal = N; Analysis: Type = Mixture; Starts=0 Model: %overall% i s| OUTCOME1@0 OUTCOME2@1 OUTCOME3@2 %C#1% [N#1@4.884]; [N#2@2.482]; i s| OUTCOME1@0 OUTCOME2@1 OUTCOME3@2 %C#2% [N#1@3.830]; [N#2@6.000]; i s| OUTCOME1@0 OUTCOME2@1 OUTCOME3@2 ......... *** ERROR The number of observations is 0. Check your data and format statement. Data file: test_PROB.dat *** ERROR Invalid symbol in data file: "*" at record #: 1, field #: 6 |
|
Jon Heron posted on Tuesday, February 06, 2018 - 6:33 am
|
|
|
Looks like a problem with the datafile you've made, rather than your model but that aside, I'm pretty sure you can only declare a growth model within the %overall% section. Jon p.s. we get told off for having a question than spans multiple posts |
|
Lydia Zhang posted on Tuesday, February 06, 2018 - 6:58 am
|
|
|
Thank you! But the data works well in the first step. It seems something is wrong with the generated new data. I checked it in other software. It looks fine though... |
|
Jon Heron posted on Tuesday, February 06, 2018 - 7:40 am
|
|
|
just had another look, it's referring to a symbol "*" which is the default symbol for missing data. it doesn't look like you are telling Mplus to look out for missing data. |
|
Lydia Zhang posted on Tuesday, February 06, 2018 - 8:01 am
|
|
|
Yes, I found the Mplus automatically changed the -9999 into * in the newly generated data. I am wondering how to tell Mplus that the missing values are * now? Should it be the following? Missing = * |
|
|
See the User's Guide. It's on our website. |
|
Lydia Zhang posted on Tuesday, February 06, 2018 - 4:46 pm
|
|
|
Thank you Dr. Muthen! I have a following up question about the manual 3-step approach. I conducted model test to see if the parameters in each classes(y on x) are significantly different. But there is nothing in the output. I am wondering if there is anything wrong in my input? VARIABLE: Names are Y X N; Missing are all(-9999); Usevar=Y X N; Classes = C(3); nominal = N Analysis: Type = Mixture; Starts=0 Model: %overall% Y ON X %C#1% [N#1@4.884]; [N#2@2.482]; Y on X(m1); %C#2% [N#1@3.830]; [N#2@6.000]; Y on X(m2); %C#3% [N#1@0.086]; [N#2@-2.441]; Y on X(m3); MODEL TEST: 0=m1-m2; 0=m2-m3 0=m1-m3 |
|
|
Please send your output to Support along with your license number. |
|
Back to top |