PLOT 3 LCGA with covariate
Message/Author
 Christine Bobbel posted on Sunday, March 17, 2013 - 6:15 am
Hello,

I got the following problem with an LCGA (4 classes and one covariate): The output is saying that my slope for class 1 is negative (-2.627), but the graph (estimated means)for this class is showing a positive slope. I don't understand how this can be?

I am an absolute beginner working with Mplus and it would be wonderful if you could help me. Thank you very much!!
 Linda K. Muthen posted on Sunday, March 17, 2013 - 7:48 am
The graph plots the means. In your model with a covariate, the intercepts are estimated not the means.
 Christine Bobbel posted on Sunday, March 17, 2013 - 8:48 am
Ok, thank you so much. Just to make sure I got it right:

This output shows the intercepts (NOT the means as in a model without covariate):

Intercepts

I 22.007 2.332 9.436 0.000
S -2.421 0.781 -3.101 0.002

And this one the effects for the regression of the intercept and slope on the covariate:

I ON
BAGE -0.009 0.019 -0.466 0.641

S ON
BAGE 0.027 0.008 3.446 0.001
 Linda K. Muthen posted on Sunday, March 17, 2013 - 8:54 am
In a conditional model, a model with covariates, the intercepts of the intercept and slope growth factors are estimated. In an unconditional model, a model without covariates, the means of the intercept and slope growth factors are estimated.
 TJ posted on Wednesday, January 01, 2020 - 12:33 pm
I would like to clarify what is the function of "specify value for variable" when plotting graphs with covariates? I am trying to plot seperate graphs for Males and Females. I checked the User Guide but it does not give detailed explaination about the function of each button. Are there resources that can help me better understand what I should be entering into those boxes when trying to plot my graphs with covariates?
 Bengt O. Muthen posted on Thursday, January 02, 2020 - 12:11 pm
If you have a variable scored 0/1 for male/female, you simply enter 0 when you want the male plot and 1 when you want the female plot.
 TJ posted on Thursday, January 02, 2020 - 7:51 pm
Thanks. And how do we obtain the Intercept and Means for the non-reference group because the intercept and mean reflected in the output file are for the reference group.
 Bengt O. Muthen posted on Friday, January 03, 2020 - 9:41 am
Perhaps you are asking how you get the mean/intercept for the non-reference group. If so, you can use Model constraint to express that and get the estimate and its SE. For instance, in the regression

Y = a + b*X + ...

X where X is scored 0/1 as above, the intercept for females is expressed in Model Constraint as

intfem = a + b;

where a and b are parameter labels defined in the Model command. See the UG for more details.
 TJ posted on Sunday, January 05, 2020 - 9:06 pm
I dont think i have b in my model command infront of gender. I only want to find out the intercept for the non-reference group male.

0=female, 1=male

Model:
I S | get@1 get@2 get@3;
S1@0;
I S ON gender;

Based on your response above, I tried adding a model constraint command:

Model:
I S | get@1 get@2 get@3;
S1@0;
I S ON gender;

Model Constraint:
new(intmale);
I S on intmale;

I was prompted with an error message saying that I need a parameter before intmale which I don't have one. does it have to do with me constraining my slope to 0?
 Bengt O. Muthen posted on Monday, January 06, 2020 - 4:01 pm
When you say e.g.

I ON Gender;

this estimates the linear regression

I = a + b*Gender + e

so you can give the label b as

I ON Gender (b);

and then use that b in Model Constraint.
 TJ posted on Tuesday, January 07, 2020 - 4:30 pm
I managed to generate an output with the following command:

MODEL:
I1 S1 | get1@0 get2@1 get3@2;
I1 S1 ON gender (b);

MODEL CONSTRAINT:
NEW (intfemale);
intfemale = b;

The output gave me the following:

INTFEMAL -0.054 0.566 -0.096 0.924

I was given the estimate -.054 and am wondering what is this estimate?
 Bengt O. Muthen posted on Tuesday, January 07, 2020 - 4:55 pm
Remember my earlier message where I said if you write the regression as

Y = a + b*X + ...

X where X is scored 0/1 as above, the intercept for females is expressed in Model Constraint as

intfem = a + b;

The label a in the Model command is given as the intercept

[y] (a);

You may want to study our RMA book where regression matters like this are discussed with Mplus inputs.