I performed a Latent Class Analysis, in order to find specific organisational structures in pre-schools. I came up with a Latent Class IV that I want to use on longitudinal vocabulary data. Specifically, I want to see if the Latent Class variable has an effect on the growth of pupilsí vocabulary. The dataset has a nested structure (1=individuals with their time measurements, 2=classrooms, 3=pre-schools), however, I want to avoid using a multilevel model. Iíve been reading that I can use type=complex, to account for the nested data structure. My question is, is this enough or do I have to use type=threelevel?
Check the intraclass correlations as a first step using Type = threelevel basic. Perhaps the pre-school variation is not large in which case you can do it as a single-level, wide format growth mixture model. Or, you can use Type=Complex to handle the pre-schools. Or, if you have few pre-schools you can handle them via dummy covariates.
Hello Dr. Muthen and thank you so much for your response.
I'd already done that, under the guidelines of Geiser, but with the twolevel option; it didn't cross my mind to do it with the three-level option. But in the end I got the same results, and I also calculated the design effect, which is larger than two.
"Average cluster size for CENTERID level 7.426
Estimated Intraclass Correlations for the Y Variables for CENTERID level
WT1 0.154 WT2 0.146 WT3 0.146 WT4 0.201"
My question is: You indicated I can use type=complex, even if I have intra-class correlation. In the manual, you are stating I have to do it in conjunction with some other options. Which are these options in my case, if I only want to use a LCGM?
Can I use the general instructions of a typical LCGM if I use type=complex, or do I have to specify cluster, between, within, center etc.?
I am new to MPLUS. I want to run a LTA model of longitudinal data were I have specified 3 classes of people at each of the timepoints (3) based on 3 variables. I want now to test the effect of those transitions in an outcome (SDQ) at timepoint 3
Here is my model notation: how can i do the above? thanks -----------
MODEL: %Overall% C2#1 on C1#1; C2#1 on C1#2; C2#2 on C1#1; C2#2 on C1#2;
C3#1 on C2#1; C3#1 on C2#2; C3#2 on C2#1; C3#2 on C2#2;
C3#1 on C1#1; C3#1 on C1#2; C3#2 on C1#1; C3#2 on C1#2; MODEL C1: %C1#1% [MVPA1]; [SLEEP1]; [SEDSC1]; %C1#2% [MVPA1]; [SLEEP1]; [SEDSC1]; %C1#3% [MVPA1]; [SLEEP1]; [SEDSC1]; MODEL C2: (equal C1 but for C2 and C3)
Prof. Beng, thanks for your reply. Actually there are two different parts in my research. I would like to start with a different one. I am running a LCGM to determine a) latent trajectories of screen time (sedsc) from age 0-4 (3 waves for 0, 2 and 4 years old). And I want to b) predict the resultant latent trajectories based on baseline SES (seifa) and gender(female). Also, I want to use a distal outcome (SDQ) at wave 3 (last timepoint of follow up) of the adjusted trajectories. Here is my notation:
VARIABLE: NAMES ARE (names) USEVAR= sedsc1 sedsc2 sedsc3 SDQ3 SEIFA1 FEMALE1; IDVARIABLE IS ID; CLASSES ARE C(3); USEOBSERVATIONS ARE (cohort EQ 1); MISSING ARE ALL (99); AUXILIARY IS SDQ3(BCH); ---tbc
ANALYSIS: TYPE = MIXTURE; STARTS = 40 8; MODEL: %OVERALL% i s | sedsc1@0sedsc2@2sedsc3@4 ; i-s@0; i s ON SEIFA1; c#1 ON SEIFA1; c#2 ON SEIFA1; i s ON FEMALE1; c#1 ON FEMALE1; c#2 ON FEMALE1; OUTPUT: SAMPSTAT STANDARDIZED CINTERVAL TECH1 TECH8; PLOT: SERIES IS sedsc1-sedsc3(S); TYPE IS PLOT3; SAVEDATA: file IS "lsac.csv"; SAVE IS CPROB;
my questions are: 1- is the syntax correct ? given my research questions above? ( Borja Del Pozo Cruz posted on Thursday, April 26, 2018 - 5:36 pm) 2- what output should i will be looking at for each of the research question? 3- if syntax is not correct, what would be a correct one for my case?