|
Abstract:
|
In a world of big data, non-probability samples are fast and easy to collect than the traditional probability samples. The main issue of big data sources is the unknown probability of inclusion, and often the case that probability is related to the response variable. To deal with the nonignorable nonresponses, we propose a double logistic regression model which linked by the response variable, and fit by MCMC methods. We compare the nonignorable model we proposed in a simulation study with a ignorable model -- single logistic regression , Horvitz-Thompson and Hajek estimators. We show that our model is favorable in terms of prediction precision.
|
