von Statistik123 » Fr 27. Jan 2023, 13:40
Hallo, alles klar, kein Problem. Jedoch meinte mein Professor, dass ich überhaupt keine Mehrebenenanalyse durchführen soll, sondern stattdessen eben die clustered standard errors. Das berühmt berüchtigte ChatGPT sagt mir Folgendes (ähnlich zu deinem Vorschlag), könnte das stimmen?:
"The "LINMIXED" command in SPSS allows you to fit linear mixed models, which can account for non-independence of observations due to clustering.
Here is an example of how you would use the "LINMIXED" command to compute clustered standard errors:
Open SPSS and enter your data
Select "Analyze" and then select "Mixed Models" and "Linear"
In the "Fixed Effects" tab, enter the dependent variable and the independent variables you want to include in the model.
In the "Random Effects" tab, select the variable you want to use as the cluster variable.
Click "OK" to run the analysis and generate the results, which will include the clustered standard errors.
Note that the syntax for the LINMIXED command is slightly different than the GLM command."
Ein anderer Vorschlag ist:
"You can use the following syntax to compute clustered standard errors:
Open SPSS and enter your data
Select "File" and then select "New" and "Syntax"
Use the "GLM" command to specify your model, and include the "CLUSTER" option followed by the variable you want to use as the cluster variable.
Here is an example of how to use the GLM command with the CLUSTER option:
GLM dependent_variable independent_variable1 independent_variable2 /WSFACTOR=cluster_variable /METHOD=SSTYPE(3)
This will run the GLM analysis and compute the clustered standard errors using the variable "cluster_variable" as the clustering variable."
Viele Grüße