A proper way to handle LPA variables PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Yu-Chung Su posted on Thursday, April 16, 2020 - 1:25 am
Hi, Dr. Mˆythen
I am conducting an LPA combining neuropsychiatric symptoms and cognitive domain and I have a few questions about the scale of the variable.
I am planning to use the severity score of the Neuropsychiatric Inventory (NPI). The NPI is a simplified clinical measure of dementia related psychiatric symptoms in 12 domains range from 0-3 (0 means no symptom).

Q1. If I include some of the domains, should I treat them simply as count variables, or should I treat them as censored variables (there are many 0 responses)?

Q2. If I use the sum of the severity scores of the 12 domain, should I treat them as continuous variables or should I treat them as censored variables?

Thank you for your time
 Yu-Chung Su posted on Thursday, April 16, 2020 - 1:28 am
Hi, Dr. Mˆythen,
I have some other quesitons.

I used various teats to measure the cognitive functions, I initially view the participants as the population and then calculate their z scores and composite scores.
In order to expand the variety of cognitive functions and the tests, I am planning to include some tests from the Wechsler Adult Intelligence Scale (WAIS) which I only have the scale scores of the test (no access to the raw scores).

Q3. Can I include the scale scores of WAIS into the raw scores of other tests to calculate their z scores and composite scores?
Because the scale score of WAIS is a kind of age-adjusted scale, I am worried if I use the results of LPA to run a logistic regression which involves the age variable would cause some related problems.

Thank you for your time and I hope you and your family are safe amongst the coronavirus pandemic!
 Bengt O. Muthen posted on Thursday, April 16, 2020 - 4:19 pm
I think these general analysis strategy questions are suitable for SEMNET.
 Yu-Chung Su posted on Thursday, April 16, 2020 - 9:00 pm
Hi Dr. Muthen,
Thank you for your advice, I have posted on SEMNET.

Some of my indicators are continuous and some are categorical, I tried to plot them and the errors showed:

*** ERROR in PLOT command
Time points for process 1 are not all continuous, all categorical, or
all latent as they should be.

Then I follow your advice in another discussion:
SERIES = cont1 (0) cont2 (1) cont3 (3) | cat1(0) cat2(1) cat3(2);

Below is my syntax:

USEVARIABLE ARE
dep_bl mem_bl ef_bl ca_bl;
CLASSES = C (2);
CATEGORICAL = dep_bl;
PLOT:
Type is PLOT3
;
SERIES IS dep_bl(0) | mem_bl(0) ef_bl(1) ca_bl(2)
;

However, I still can see the plot of the categorical varibles profile of different class.

Is there any way that I can view the profile of the class containg both continuous and categorical variable? What should I do to achevied this goal?
 Bengt O. Muthen posted on Friday, April 17, 2020 - 6:17 am
When you have a combination of continuous and categorical outcomes, a plot is not provided by Mplus because the y-axis scale is different for those two types: Continuous outcomes have means and categorical outcomes have probabilities. You can do your own plot in some way that you find useful. The output results give the means and the probabilities that you would use.
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: