I’m having trouble generating out-of-sample predictions for a simple linear model with missing covariate data. I understand how to fit a model with missing covariate data using a numpy masked array: import numpy as np import pymc3 as pm # generate data with missing values X_train = np.random.rand…

Not sure if it is the best way, but my solution would be the following: 1, build a model that allows having missing values in the covariates (i.e., the predictor X). This is similar to your model one. However, to have a more meaningful prediction for missing input, a more informative prior is neede…

OOS predictions with missing input values

Questions

aayushy November 22, 2018, 10:52am 7

I know this is a really old thread but can someone please help me figure out how I can use PyMC3 to do exactly as quoted?

Topic		Replies	Views
Can't get sample_prior_predictive to work with missing values Questions	4	769	June 10, 2021
Out of sample predictions from a pickled model Questions theano	14	2728	January 31, 2019
How to predict new values on hold-out data Questions	24	13494	July 22, 2020
Logistic Regression w/ Missing Data? Questions	7	2901	September 11, 2017
OOS predictions of covariate with "out-of-range" missing values Questions	6	672	August 12, 2024

OOS predictions with missing input values

Related topics