Improper Solutions in Mplus

preview_player
Показать описание
Improper solutions (e.g., negative variance estimates; Heywood cases) occur frequently in confirmatory factor analysis (CFA) and structural equation modeling (SEM). QuantFish instructor Dr. Christian Geiser shows how improper solutions can be detected in the Mplus program, what possible causes are, and how this issue can be addressed.

#Mplus #Mplusforbeginners #statistics #CFA #SEM

Dr. Geiser's BOOKS:

ON-DEMAND MPLUS COURSES:

GET IN TOUCH with DR. GEISER:

ABOUT THIS INSTRUCTOR:
Dr. Christian Geiser is a quantitative psychologist, author of two books on Mplus, and a leader in the development of latent variable techniques for complex data. With his accessible books and sought-after workshops, he has helped thousands of researchers and students around the world to achieve their analytic goals.

ABOUT QUANTFISH:

Some affiliate links may be included at no additional cost to purchasers.
Рекомендации по теме
Комментарии
Автор

Dear Christian,
(I'm sorry for my English)
Thanks for these great videos. With this, plus the help of my thesis supervisors, I was able to solve a problem that had taken me several days. By the way, I have one of your books on MPlus. Thanks again.
Best regards.
Flavio.

fmunozt
Автор

Dear Christian,
Thank you very much for making this Video! Can you make another one to explain another case of Heywood cases? That is, factor correlations > 1 in Mplus. Please also propose your solutions for this issue.

邵凯祺
Автор

Dear Christian.

Thanks for this video. I saw it several times and now something similar happened to me.
Can you give me your opinion according to your experience, please?
I work with a correctly specified CFA model, however, with a sample from another country, there is a negative variance in one factor. I have already tried everything recommended by my thesis directors, without good results. I accidentally set the variance of a latent variable (one that shows negative variance) to a value between 0 and 1 (not 1, as recommended). Then the model fitted.
Would this make any sense, any theoretical support? I just think that setting a variance to 1 is arbitrary. If the problem is that it produces negative variance, why not arbitrarily set it to a value less than 1?

My model in R is:
model1 <- 'F1 =~ V03+V08+V13+V18
F2 =~V01+V06+V11+V16
F3 =~ V02+V07+V12+V17
F4 =~ V04+V09+V14+V19
F5 =~V05+V10+V15+V20
G1 =~ F1+F2+F3
G2 =~ F3+F4+F5 #The original model ends here.
F3 ~~ 0.501*F3 #Here begins my strange procedure
F4 ~~ 0.167*F4
F5 ~~ 0.163*F5'

Thank you very much and sorry for my English.

Tour follower,
Flavio Munoz

fmunozt