Beta parameter with huge standard error

[mspinola]

Dear forum members,

When there is a beta estimate that has a huge se, like the 4th parameter below, is it mean that the real parameter estimate is not reliable?


Label Estimate SE LCI UCI
p -1.45921 0.4565816 -2.35411 -0.56431
p -0.3022797 0.2262076 -0.7456467 0.1410872
Psi 0.1211712 0.7194736 -1.2889971 1.5313395
Psi 14.07518 3247.7838 -6351.5813 6379.7316

The number of parameters in the MARK Result Browser is 3 (instead of 4).
Thank you very much in advance.
Best,

Manuel Spínola

[darryl]

Hi Manuel
Depends what the 4th parameter relates to...
Darryl

[darryl]

If 4th parameter relates to the effect of a categorical covariate then this suggests the estimated probability for that category is essentially 1. As this is on the boundary of allowable values for a probability, it can create some issues and problems with the estimated SE. However, the SE for the real parameter is probably 0. I suspect this is a real estimate though so the parameter count should be 4 not 3