The causal effects of continuous treatments can be examined using generalized propensity scores for removing selection bias. In the video below, I demonstrate the estimation of generalized propensity scores and the dose response function using the R software. The example is to estimate the effects of logins in the Algebra Nation Virtual Learning Environment per school on the school's mean Algebra 1 End of Course Exam scores.
Another approach to remove selection bias in the estimation of the average treatment effect of continuous treatments is to use inverse probability weights. In the video below, I show how to estimate the weights using generalized linear models, the covariate balancing propensity score (CBPS) method, and Bayesian additive regression trees (BART).
The success of bias removal using inverse probability weights for a continuous treatment can be evaluated with measures of covariate balance. In the video below I show how to evaluate covariate balance for inverse probability of treatment weighting of continuous treatment doses.
In this final video I show how to estimate the average treatment effect of a continuous treatment does using inverse probability weighting. I also show how to perform a sensitivity analysis.
R code for Propensity Score Analysis for Continuous Treatments
chapter7_part1_generalized_propensity_score_and_dose_response_function.r  
File Size:  7 kb 
File Type:  r 
chapter7_part1b_generalized_propensity_score_and_dose_response_function_with_gamma_distribution.r  
File Size:  7 kb 
File Type:  r 
chapter7_part2_inverse_probability_of_treatment_weights_for_continuous_treatments.r  
File Size:  4 kb 
File Type:  r 
chapter7_part3_evaluation_of_covariate_balance_for_continuous_treatments.r  
File Size:  4 kb 
File Type:  r 
chapter7_part4_estimation_of_the_average_treatment_effect_for_continous_treatments.r  
File Size:  3 kb 
File Type:  r 
Example Data for Propensity Score Analysis for Continuous Treatments


Proudly powered by Weebly