Hi, I am trying to solve a silly indexing issue and realize I’m not sure I understand pymc3 syntax. Suppose I want to model a y that is normally distributed, and have predictors x0 and x1. x1 has only sometimes been measured, but I do not wish to impute it. Instead, I want to define two likelihoods…

That looks fine. (At least pymc syntax wise, I really don’t know about the imputation/non-imputation). There is no need to use theano shared variables, that is only something you can use to switch out datasets. Shouldn’t the second sigma be something like tt.sqrt(sigma**2 + sigma_m**2)?

Two likelihoods dependent on missing predictors

Questions

jsevo August 22, 2019, 7:59pm 3

Thanks! And you are right.

Topic		Replies	Views
What does the hierarchical model look like when having missing in observed? Questions	12	1173	October 31, 2018
Calculating the likelihood based on not missing observed values Questions	5	1995	June 12, 2018
Measurement Uncertainties in Multiple Regression Questions	3	1457	April 28, 2020
Masking missing values of predictors Questions	3	1385	July 10, 2020
OOS predictions with missing input values Questions	6	4140	November 22, 2018

Two likelihoods dependent on missing predictors

Related topics