Ordinal response different coeffecient for each response cat

[frank]

Hi,
I followed the steps from chapter 11 and set up a model for my parent-satisfaction-study.
Now I would love to know how to decide wether the differences between coeffecient are considered small (like on page 164 of the 2008 manual) or not.

My lowest level of satisfaction has a little number of cases.
very unsatisfied - 23 parents
rather unsatisfied -133 parents
rather satisfied -705 parents
very satisfied - 259 parents

If I let the coeffecients vary for each responsecatergory, for some variables the coeffecient are not significant for the "very unsatisfied category" (using a Wald-Test).
Is this normal with such a small number of cases in this category?
What should I do?
Thanks for your help
Best
Frank J.

[frank]

Can anybody please help me with this?
Or tell me where I could find something on this in the literature?
Thank you very much.

[frank]

Can please someone explain how to test if the odds are proportional, like shown here: http://www.cmm.bristol.ac.uk/MLwiN/tech ... models.pdf on page 14.
How do I decide if thats the case?