|
|
 |
| Thresholds with missing data |
 
|
| Message/Author |
|
|
|
|
I have ordered categorical data with 5 response options. I have missing data. In the descriptives, Mplus says I have 6 categories, which I assume is Mplus treating missingness as a "category." However, in the CFA model, Mplus is estimating 5 thresholds per variable rather than 4. With 5 ordered categories, it is my understanding there should only be 4 thresholds per variable. I am worried Mplus is treating missingness as a 6th "category" and including that in the model.
Here are select lines of my input:
usevar
V_app1 V_app2 V_app3 V_app4 V_app5;
missing are all (-9999);
categorical are all;
analysis: type = general missing;
estimator = WLSMV;
model:
app by
V_app1 V_app2 V_app3 V_app4 V_app5;
Here are selected lines of my output:
V_APP1
Category 1 0.004 2.000
Category 2 0.042 22.000
Category 3 0.117 62.000
Category 4 0.253 134.000
Category 5 0.287 152.000
Category 6 0.298 158.000
! I assume Category 1 is the missingness
Thresholds
V_APP1$1 -2.672 0.237 -11.280 0.000
V_APP1$2 -1.692 0.095 -17.852 0.000
V_APP1$3 -0.985 0.065 -15.106 0.000
V_APP1$4 -0.214 0.055 -3.907 0.000
V_APP1$5 0.530 0.057 9.245 0.000
! I am worried the first threshold ($1) is connecting my missingness "category" to my first real data category |
|
|
|
|
| No, if you have specified a missing data flag (like -9999), then missing data is not a category. There must be other values in the data. If you can't find the problem you can send your output and data to Support along with your license number. |
|
|
| Back to top |
|
|
