Interpreting thresholds (in version 5...
Message/Author
 Johnny Wu posted on Wednesday, March 12, 2008 - 12:31 pm
Hi Dr Muthen, I have a Multiple categorical indicator latent growth model. Three binary indicators per time point, three time points, and two growth factors (i and s).

"sub" is my substance factor
"tob" is binary tobacco use
"alc" is binary alcohol use
"dru" is binary other drug use

MODEL:

sub8 BY tob8 (1);
sub8 BY alc8 (2);
sub8 BY dru8 (3);

sub10 BY tob10 (1);
sub10 BY alc10 (2);
sub10 BY dru10 (3);

sub12 BY tob12 (1);
sub12 BY alc12 (2);
sub12 BY dru12 (3);

[tob8\$1] (4);
[tob10\$1] (4);
[tob12\$1] (4);

[alc8\$1] (5);
[alc10\$1] (5);
[alc12\$1] (5);

[dru8\$1] (6);
[dru10\$1] (6);
[dru12\$1] (6);

isub ssub |

sub8 @0
sub10 @1
sub12 @2
;

[isub];
[ssub];

isub;
ssub;

I know that thresholds to binary outcomes are analogous with means to continuous outcomes. So what is the equation to convert the "threshold" to a meaningful number(i.e., a number between 0 and 1)? My output reads:

THRESHOLDS:

tob8\$1 estimate = 1.132

alc8\$1 estimate = 1.413

dru8\$1 estimate = 3.099
 Linda K. Muthen posted on Wednesday, March 12, 2008 - 6:04 pm
Thresholds are not particularly interesting parameters. You could turn them into probabilities but if you know the observed proportions, that would not give you new information.