www.cmm.bristol.ac.uk/forum

Hi,
maybe it's not the right place to ask but if it isn't maybe someone can tell me where to do.
I 've gathered data on the satisfaction of parents )(n=1195) with their kidsprimary school. The data are hierarchical (8 to 24 parents from 2 classes eachfrom 41 schools).
The problem is that the variables that I want to explain are only ordinal scale(very satisfied ... very unsatisfied).
I have some more detailed views on the satisfaction with school/class withother items (which are also on an ordinal scale (strongly agree stronglydisagree).
I have data on the school/class-situation I want to use to explain thesatisfaction of the parents.
Is there a way to analyze this kind of data, keeping the hierarchical structurein mind even though the response variable isn't metric?
Do you know any studies (maybe attitude-studies) or papers who dealt with thisproblem?
Thank you for your help.
Best
Frank J.

Yes, you can certainly use multilevel models with this kind of response. Chapter 11 in the MLwiN User's Guide takes you through the process. You probably want to be using MCMC with this kind of model, so you might want to read the relevant chapter in the MCMC manual as well. There are also some slides on this topic on the CMM website's workshops page: http://www.cmm.bristol.ac.uk/MLwiN/tech ... ndex.shtml.

Thank you for your reply. I will look into it the next days.
Have some nice holidays
Frank J.

Hi,
I tried to work as described in chapter 11 and I got told I should care about my missing values as no missing values are allowed in this kind of modells.
Now I am kind of stuck as I wonder how to deal with this.
I got 1200 parents from 76 classes.
If I included all my predictors the number of parents goes (due to missing values) down to 740.
That seems a lot to me and maybe it doesnt make sense to start deleting all these cases listwise build a modell and then start over because its to much loss of cases.
But I currently know very little about imputation.
Is there a good way to start (reading) on this topic?
Thank you for your help
Frank J.

Hi,
I also dont know how to delete cases with missing data.
I tried the Listwise menu but I dont know the value (in the data view MISSING is shown but missing dont work).
So is there a way to delete all cases with missing values?
Thx for your help

You should be able to choose the missing value in the "Listwise" screen by clicking in the box to the right of "Listwise delete on value" and pressing the "m" key. Did you try this, and if so what was the error message/problem that you got?

I had the same question. Thank you for the information, much appreciated.