# Warning on ordinal regression

#### frank100

##### New Member
I have a dependent variable:low,medium,high (coded as 1,2,3) and
10 independent continuous variable and 190 cases.

When I run ordinal regression(PLUM) every things seems fine expect I get this warning at the top.
Warnings
There are 315 (63.5%) cells (i.e., dependent variable levels by observed combinations of predictor variable values) with zero frequencies.

I tried online but could not find any concrete answer.

Code:
Model Fitting Information
Model	-2 Log Likelihood	Chi-Square	df	Sig.
Final	       248.139	            82.702	10	.000
Link function: Logit.
Goodness-of-Fit
Code:
	Chi-Square	df	Sig.
Pearson   432.116	     334	.010
Deviance	248.139	334	1.000
Link function: Logit.
What should i do?
I am trying to test hypothesis on effect of some of the predictor on dependent variable.

If I use multiple regression every thing looks fine but i don't think i can do that.

#### hlsmith

##### Not a robit
Is the issue that you have a high-dimensional space, given many variables, and sparsity in the subgroups so the model has quasi or complete separation issues? For example a subgroup (males, old, certain race, certain condition,...) with null count for the number in the category.

#### frank100

##### New Member
There is no way I can conduct more survey and add more data(and from the analysis it seems like I will need around 1000 cases) so I am planning to convert dependent binary variable as binary and carry other regression. May be that will help.