Bounds on estimates
Message/Author
 Eric Turkheimer posted on Wednesday, June 09, 2004 - 12:00 pm
What is available to place bounds on estimates? I have variances that keep going negative. I can think of tricks to accomplish it by forcing Mplus to square the parameter, but I was wondering if there might be an easier way.

Thanks,
Eric
 Linda K. Muthen posted on Thursday, June 10, 2004 - 8:37 am
We usually recommend fixing small negative variances to zero and if they are not small rethinking the model.
 Anonymous posted on Monday, October 25, 2004 - 9:29 am
Is there any way to bound parmeter estimates between two values. I would like to restrict factor loadings to values between -4 and 4
 bmuthen posted on Monday, October 25, 2004 - 3:49 pm
Yes, through a trick using Model Constraint. Tihomir suggests using the two facts that the sine function is bounded by -1, +1 and that we can work with a dummy parameter (called p1 below).

For instance, in a factor model, let parameter p2 be the one to be bounded:

f by y1 y2(p2);
!create a dummy parameter p1 (any one will do):
fdum by y1@0; fdum (p1);

Model Constraint:
p2 = 4*sin(p1);

The estimation will then take place with respect to p1 and p2 will be the bounded factor loading.
 MAH posted on Thursday, August 07, 2008 - 8:45 am
Similar question with respect to boundary constraints. I want to constrain an estimate to be between 0-1. Is there an easy way to do this in Mplus 5.1?
 Bengt O. Muthen posted on Friday, August 08, 2008 - 9:18 am
You can use labeling of Model parameters and then in Model Constraint say

p>0; p<1;