Estimating Marginal Versus Conditional Parameters
If your goal is the estimation of an effect, many people who study causal inference prefer the estimation of marginal parameters (such as the average treatment effect and marginal structural models) to the estimation of conditional parameters.
Many have asked me whether there is a practical difference between the two. I outline my views and provide some numerical comparisons in this pdf.
Many have asked me whether there is a practical difference between the two. I outline my views and provide some numerical comparisons in this pdf.
| marginal_versus_conditional_effects.pdf |
The numerical comparisons in the pdf were done in Wolfram Mathematica and checked in version 11.
| biasplots.nb |
