Question

True or False: In the case of a completely randomized experiment with a large sample (with negligible selection bias), the treatment effect on the treated E[YTi|T]–E[YCi|T] only reflects the effect for the units assigned to be treated, but cannot speak to the treatment effect for the control group or the larger population.

Answer 1

FALSE. The entire point behind an experiment is to draw inferences for the population based on a test on sample. Here, as we see that there were negligible selection bias and the experiment was completely randomized. This implies that there were no biasness in selecting the control or the treatment groups. Thus, the units that actually fell into the control group had equal chances of being considered under the treatment group.

Hence, the results should be generalizable to the population (and hence also the control group)