Questions about LPA PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Solah Lee posted on Friday, July 10, 2020 - 10:11 pm
Q. Below is Syntax I used for LPA analysis.
-----------------------------------------
TITLE: LPA_CLASS(5)
DATA: FILE = aloneness.dat;
VARIABLE: NAMES = id sex age job edu resi relation term lon avoida anxia sds nsds
confid sregul taskd facp distp;
USEVARIABLES = lon avoida anxia sds nsds;
CLASSES = C(5);
ANALYSIS: TYPE = MIXTURE;
LRTBOOTSTRAP =100;
starts = 100 10;
stiterations = 20;
OUTPUT: TECH11 TECH14;
-----------------------------------------
Among “VARIABLE: NAMES”, the variables used for LPA are <lon>. And these variables(sex edge job edu resi relation term (excluding id)) contains demographic information for each individual.

When the number of groups is determined through LPA analysis, I would like to know the demographic information of the people in each group. Is there any way?
[ex. class 1(total sex = 90 people): sex 1 = 60 people, sex 2 = 30 people,/ class 2(total sex = 75 people): sex: 1 people,= 35, sex = 40 people]

I wonder if I just couldn’t find this information although it is already provided with LPA result or if I need to add another command along with the above syntax.
 Bengt O. Muthen posted on Sunday, July 12, 2020 - 4:43 pm
You can use the Most Likely Class classification to look at the demographics. Or you can let the demographic variables predict class membership - then you get the information directly by TECH7.
 Solah Lee posted on Monday, July 13, 2020 - 8:29 pm
Q-1. Is “Most Likely Class classification” the basic information that comes out when I enter and analyze the Syntax that I uploaded before? I can't find this data in the Output file I analyzed.


Q-2. I specified five variables(lon avoida anxia sds nsds) as usevariable and added TECH7 to Ouput(Output: TECH7 TECH11 TECH14;) and analyzed the file. However, the TECH7 output showed some information only about these five variables(on avoida anxia sds nsds) not about demographic information.

I don't want any other variables than those used for LPA analysis(lon avoida anxia sds nsds) to affect profile formation, and I want to check the demographic information(sex age job edu resi relation term) of individuals in each group classified through these five variables(lon avoida anxia sds nsds).
[ex. class 1(total sex = 90 people): sex 1 = 60 people, sex 2 = 30 people,/ class 2(total sex = 75 people): sex: 1 people,= 35, sex = 40 people]

How can I get the information I need using TECH7? I don’t want the demographic information to affect the LPA groups. I couldn’t find any information about how to using TECH7, so I am adding another question.

Thank you so much!
 Tihomir Asparouhov posted on Tuesday, July 14, 2020 - 6:33 pm
Use

SAVEDATA: FILE=1.dat; SAVE=CPROB;

That will give you the most likely class variable in the file as well as the auxiliary variables. You can then tabulate it in whichever way you want.
 Solah Lee posted on Tuesday, July 14, 2020 - 11:08 pm
I have added SAVEDATA: FILE=1.dat; SAVE=CPROB; and “1.dat” file has been created.

There are numbers listed in the “1.dat”file, like the "aloneness.dat" file I used for analysis.

Can't I see the demographics statistics of each group? Do I have to count the demographic information by referring to the probability information for each group in this file?
 Tihomir Asparouhov posted on Wednesday, July 15, 2020 - 8:36 am
Actually you must add this option to the VARIABLE command:

auxiliary=sex age job edu;

At the end of your output file see the section
"SAVEDATA INFORMATION"

That describes the columns in the 1.dat file. After the the CPROB1 ... is the most likely latent class variable. This is the complete information for the data and the most likely class but it is not tabulated. You can use separate run to tabulate it, like Mplys type=basic runs for each group, or do it in spreadsheet software.

There is a great deal of methodology that attempts to provide these quantities while also taking care of the uncertainty in the most likely class variable. If you are interested to learn about that you should read Mplus Web notes No.15 and No.21, or if you want to see a quick glimpse at what this is about use

auxiliary=(bch) sex age job edu;

or

auxiliary=(e) sex age job edu;
 Solah Lee posted on Thursday, July 16, 2020 - 3:46 am
I was planning to use BCH command to test the difference about dependent varibles(confid sregul taskd facp distp) between groups in later analysis, and the material you provided was very helpful.

In the analysis of the questions I asked before, I want to know the sum of demographic information by each group.
[ex. class 1(total sex = 90 people): sex 1 = 60 people, sex 2 = 30 people,/ class 2(total sex = 75 people): sex: 1 people,= 35, sex = 40 people]
So I think it would be most appropriate to use Analysis: type = basic; you told me.
If this command gives me the information I need.

As the first step, I entered Syntax as below as you told.

TITLE: LPA_CLASS(3)
DATA: FILE = aloneness.dat;
VARIABLE: NAMES = id sex age job edu resi relation term
lon avoida anxia sds nsds
confid sregul taskd facp distp;
AUXILIARY = id sex age job edu resi relation term;
USEVARIABLES = lon avoida anxia sds nsds;
CLASSES = C(3);
MISSING = ALL(999);
ANALYSIS: TYPE = MIXTURE;
LRTBOOTSTRAP =100;
starts = 100 10;
stiterations = 20;
OUTPUT: TECH11 TECH14;
PLOT: TYPE = PLOT3;
SERIES = lon(1) avoida(2) anxia(3) sds(4) nsds(5);
SAVEDATA: FILE=1.dat;
SAVE=CPROB;
 Solah Lee posted on Thursday, July 16, 2020 - 3:46 am
"file 1.dat" was created and according to "SAVEDATA INFORMATION", the variable order of the generated files is "lon avoida anxia sds nsds saxage job edu resi relation term cprob1 cprob2 cprob3 c".

I understood that if I use the command “analysis: type = basic;” with the newly created "file 1.dat", I can tabulate demographic information(ex. for sex - how many men or women belong to group 1, not the average). Did I get it right?

I have problem understanding how to write Syntax to get demographic information for each group. Also, I wonder if I have to designated the demographic information as categorical(or nominal), since this demographic variables are not continuous.

I've written Syntax as I know, and looks like it is never enough for the information I need. I'm wondering what I need to revise.
I am very sorry about asking many question. It’s been really helpful.
Thank you very much!

TITLE: LPA_BASIC
DATA: FILE = 1.dat;
VARIABLE: NAMES = lon avoida anxia sds nsds
id sex age job edu resi relation term
cprob1 cprob2 cprob3 c;
NOMINAL = sex age job edu resi relation term c;
ANALYSIS: TYPE = BASIC;
 Tihomir Asparouhov posted on Thursday, July 16, 2020 - 4:53 pm
In the variable command add

USEOBSERVATION ARE (C EQ 1);

to get the results for the first group.

Also add
USEVAR = sex age job edu resi relation term;

remove

NOMINAL = sex age job edu resi relation term c;

and add

CATEGORICAL = sex job edu;

if those are the categorical variables.
 Solah Lee posted on Sunday, July 26, 2020 - 6:14 am
Hello.
I got the result I wanted.

Thank you so much for your help!
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: