@AlexAndorra @tirthasheshpatel The last doubt about my model because I’m going around it and I can’t get it. You would help me understand it much better.
My goal is get (from the posterior) the probability ([0,1]) of EACH category for specific features. (example : 4 features (day of the month of the movement == 10, times the movement is repeated in a month == 2, individual cost per movement == 50, total spend per month of each movement == 100).
What numpy.array for observed value is valid ? or am I saying something crazy? 
(in the code I generate a value of observed with 20000 lines of 1s for the category_1 and 0s for the other categories (the other 40000 lines), because I want to evaluate what I have told you before but only for the category_1)
Thankyou from a beginner!