Compute "growth" fscore and plot PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Luis Anunciação posted on Thursday, February 08, 2018 - 1:42 pm
I would like to say thanks for all your support. I didn´t say that in the previous message because of the text limit.
 Luis Anunciação posted on Thursday, February 08, 2018 - 1:42 pm
I have 2 groups (same students,two points-T1/T2). They were measured in the same 2-factor scale. My ds is in the long format (u1 u2, ..., grp)
I want to perform a growth and invariance analysis,compute fscore for each person at T1 and T2 and, after plot. T2 must consider T1.
The demo of my code:

NAMES ARE u1-u6 g;
USEVARIABLES ARE u1-u6;
CATEGORICAL ARE u1-u6;
GROUPING IS g (1=year3 2=year4);
ANALYSIS: PARAMETERIZATION = THETA;
ESTIMATOR = WLSMV;
MODEL:
f1 BY u1@1
u2 (f21)
u3 (f31);
f2 BY u4@1
u5 (f52)
u6 (f62);
[u1$1] (t1); (etc)
f1*; f2*;
f1 WITH f2*;
[f1@0];
[f2@0];
u1@1;
u2@1; (ETC)
MODEL year4:
f1 BY u1@1
u2 (f21)
u3 (f31);
f2 BY u4@1
u5 (f52)
u6 (f62);
[u1$1] (t1);
[u2$1] (t2); (ETC)
f1*; f2*;
f1 WITH f2*;
[f1*];
[f2*];
u1@1;
u2@1; (ETC)
OUTPUT: TECH1; STDYX;
savedata:
file is fscores.dat;
save=fscores;
 Bengt O. Muthen posted on Thursday, February 08, 2018 - 4:07 pm
Do you have 2 groups, or 1 group measured twice? Or 2 groups, measured twice?

If it is 1 group of the same students measured twice, a 2-group analysis is not correct, but instead you have 1 group with twice as many variables (t1, t2).
 Luis Anunciação posted on Thursday, February 08, 2018 - 4:49 pm
Just one group measured twice. The same student was measure last year and this year with the same scale.
I was following Millsap and Yun-Tein's instructions and that´s that's the why I modeled like that. Could you please link some help file.
Thanks again for all support.
 Bengt O. Muthen posted on Friday, February 09, 2018 - 5:06 pm
With a single group measured twice you would not use the Grouping option. Grouping is for independent observations which would not be the case with the same subject measured twice. Instead, use a regular single-group analysis where you analyze 2*p outcomes where p is the number of outcomes at each of the 2 time points. The measurement invariance testing can still be done (across time).
 Luis FCA posted on Saturday, February 10, 2018 - 10:37 am
Thanks, Dr. Muthen. I found some help at this forum: www.statmodel.com/discussion/messages/14/7204.html?1489094573

Since the same participants were assessed twice with the same bidimensional scale, I changed the syntax as it follows:

MODEL:
!time 1 emo
emo_t1 BY y1
y2-y12 (2-12);

!time 2 emo
emo_t2 BY y22
y23-y33 (2-12);

!time 1 soc
soc_t1 BY y13
y14-y21 (13-20);

!time 2 soc
soc_t2 BY y34
y35-y42 (13-20);


!thresholds to be the same across 2 time points
[y1$1 y22$1] (a); (etc)
...
{y1-y22@1 y13-y34};

i s | emo_t1@0 emo_t2@1 soc_t1@0 soc_t2@1;

Do you have any concerns about that?
Thanks much for all support.
 Luis FCA posted on Sunday, February 11, 2018 - 11:04 am
By the way, I just want to say thanks for publishing your class on the internet. I found a lot of help watching the "Topic 8": https://www.youtube.com/watch?v=tI2oOtZboKc&t=7005s at minute 60:07 and page 32 of https://www.statmodel.com/download/Topic%208.pdf
These materials are precious and help a lot!
 Bengt O. Muthen posted on Sunday, February 11, 2018 - 2:40 pm
Regarding your first posting, see the growth modeling parameter specification table on page 751 of the Version 8 User's Guide available on our website.
 Luis FCA posted on Monday, February 12, 2018 - 6:48 am
Truly precious!
I've got 99.9% of my answers replied in this document and your video training. Thanks for publishing all material.

The last question I have is related to plot function.
Because I want to see individual latent growth (from T1 to T2 regarding each dimension and each item), what's similar to this Mplus plot (https://stats.idre.ucla.edu/wp-content/uploads/2016/03/lcat1.gif) I specified:

TYPE = PLOT3;
SERIES is y1-y42 (s);
(each item)
or
SERIES is emo_t1 emo_t2 soc_t1 soc_t2 (s);
(each factor)

But I could not find any option for individual line plot.

Probably, this post forum is similar to my question, but I don't know if the answer still the same: Reference: Dr. Linda message on December 04, 2011 - 6:15 am
http://www.statmodel.com/discussion/messages/12/7387.html?1323048375

Thank you.
 Bengt O. Muthen posted on Monday, February 12, 2018 - 5:08 pm
Using UG ex 6.14, you get an estimated growth means plot by

Plot:
type = plot3;
series = f1-f3(s);

and then requesting Estimated means.

There is no automatic way of getting estimated growth plots for the factor indicators. You would have to express their means over time using model parameter labels in Model Constraint and use the LOOP and PLOT options.
 Luis FCA posted on Thursday, February 15, 2018 - 3:57 am
Thanks much, Dr. Muthen!
All questions are clear now!
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: