p-values logistic regression

[Simon Dirmeier]

Dear all,

when fitting a logistic model in pyspark 
(https://spark.apache.org/docs/2.2.0/ml-classification-regression.html#binomial-logistic-regression) 
in many cases, the summary does not contain p-values, or rather calling 
the summary throws an exception (even though in these cases 
/model.hasSummary/ is true). In some cases it works though.

Is that because the vcov cannot be inverted and thus no test statistics 
can be computed?  Or what might be the problem here?

Thanks,

Simon