In short, I am trying to take this functioning SEM model
and apply the following changes:
- Make the "CSykKmTransp_sqrt_cat4" treated as the ordinal variable it is, through applying ologit.
- Fix the error term of the summated attitude scores by (1-(Cronbach's alpha))*variance to account for the measurement error in this construct. The summated attitude scores are post_att_tot and pre_att_tot, and are the rowmean of 6 other variables.
To implement this, I am currently attempting to convert the model to gsem, adding ologit to the CSykKmTransp_sqrt_cat4 link, and adding the Cronbach's alpha through the reliability() option. The current code is this, but it is not working.
There are two main errors i get, that I am struggling with.
1 ) "invalid reliability() option; pre_att_tot must be an observed endogenous gaussian variable with an identity link"
and of this reliability option is removed,
2) "invalid covariance specification; pre_att_tot does not identify a gaussian error or latent variable"
Which I do not completely understand, because pre_att_tot not being endogenous was not a problem in the SEM original model. However, I realize that I am currently working with model specifications way above my current SEM knowledge and would greatly appreciate help.
Code:
qui: sem ///
(post_att_tot <- pre_att_tot ASykKmTransp_sqrt_cat4) ///
(CSykKmTransp_sqrt_cat4 <- ASykKmTransp_sqrt_cat4 pre_att_tot) ///
, nocapslatent ///
cov(pre_att_tot*ASykKmTransp_sqrt_cat4 e.post_att_tot*e.CSykKmTransp_sqrt_cat4)
sem, standardized
- Make the "CSykKmTransp_sqrt_cat4" treated as the ordinal variable it is, through applying ologit.
- Fix the error term of the summated attitude scores by (1-(Cronbach's alpha))*variance to account for the measurement error in this construct. The summated attitude scores are post_att_tot and pre_att_tot, and are the rowmean of 6 other variables.
To implement this, I am currently attempting to convert the model to gsem, adding ologit to the CSykKmTransp_sqrt_cat4 link, and adding the Cronbach's alpha through the reliability() option. The current code is this, but it is not working.
Code:
gsem ///
(post_att_tot <- pre_att_tot ASykKmTransp_sqrt_cat4) ///
(CSykKmTransp_sqrt_cat4 <- ASykKmTransp_sqrt_cat4 pre_att_tot, ologit) ///
, nocapslatent reliability(pre_att_tot 0.7469 post_att_tot 0.7492) ///
cov(pre_att_tot*ASykKmTransp_sqrt_cat4 e.post_att_tot*e.CSykKmTransp_sqrt_cat4)
1 ) "invalid reliability() option; pre_att_tot must be an observed endogenous gaussian variable with an identity link"
and of this reliability option is removed,
2) "invalid covariance specification; pre_att_tot does not identify a gaussian error or latent variable"
Which I do not completely understand, because pre_att_tot not being endogenous was not a problem in the SEM original model. However, I realize that I am currently working with model specifications way above my current SEM knowledge and would greatly appreciate help.
Code:
* Example generated by -dataex-. For more info, type help dataex
clear
input float(pre_att_tot post_att_tot ASykKmTransp_sqrt_cat4 CSykKmTransp_sqrt_cat4)
3.166667 3.833333 1 1
6.666667 6 3 3
6 5.666667 4 3
4.3333335 4.3333335 1 1
3.5 3.5 3 3
5.666667 6.333333 2 3
6 7 4 4
4.6666665 4.6666665 3 3
5.5 6 4 4
3.5 4.833333 2 2
3.166667 6 2 2
5 3.833333 4 4
4.6666665 5.166667 4 4
5.166667 4.3333335 1 1
6.5 6.5 4 4
4.833333 4.6666665 4 4
4 4.5 3 3
5.333333 5.666667 3 4
6.666667 6.666667 4 4
3.833333 3 1 2
2.5 2.1666667 1 1
6 6.666667 4 4
4.6666665 3.5 1 1
3.333333 2.3333333 1 1
5 5.333333 4 3
6 5.833333 4 4
6.166667 6.333333 4 4
5.333333 5.166667 4 4
6.166667 6.166667 4 4
6.666667 6.5 3 4
5.333333 4.833333 4 4
5.833333 5 3 4
7 6.333333 4 4
5.333333 5 3 3
5.5 6 2 2
5.833333 5.5 4 3
3.5 5.333333 3 2
6.333333 6.666667 3 3
3.166667 4.833333 4 1
5 4.5 3 1
4.5 4.3333335 4 4
5 3.333333 3 2
6.666667 6.333333 2 3
4.6666665 5 4 3
7 6 2 3
6.5 6.666667 4 4
4.833333 2.3333333 1 3
5 5.833333 4 4
5.666667 4.5 1 1
5.5 5 4 4
6.333333 6 1 3
4.1666665 4.3333335 3 3
4.6666665 5 1 1
7 7 4 4
4.833333 4.1666665 4 4
4.833333 4.833333 3 3
6 6.166667 1 1
5 7 4 4
4.6666665 5.5 4 3
5.833333 5 4 4
6 5 3 1
3.833333 5.333333 . 2
6.333333 5 4 3
4.5 4.6666665 3 3
6 5.833333 4 4
5 2.833333 3 4
4.5 3.833333 1 2
4.6666665 4.833333 1 3
5.833333 5.666667 2 2
6.833333 6.5 2 3
3.5 3.333333 1 3
5 6.333333 4 3
7 5 1 1
6.666667 5.833333 3 4
4.5 3.5 1 1
5 5.166667 2 2
4.1666665 4 3 2
5.5 5.833333 4 4
4 4.833333 1 1
5.5 5 1 2
5.833333 7 3 4
5 5.5 4 4
5 6.5 3 2
3.666667 4.3333335 4 4
5 5.166667 3 3
4.5 4.833333 3 3
5.166667 5 1 1
5.5 6.5 2 3
5.666667 6 3 4
7 7 4 4
5.333333 5.666667 1 2
6.5 7 4 4
2.833333 4 2 4
4.833333 4.833333 3 3
4.3333335 4.5 3 3
6 7 4 4
5.666667 5.5 3 4
5.333333 4.5 4 4
3.5 4.6666665 1 1
3.666667 4.5 2 2
end