Natural experiment on the effect of a binary treatment on viewing metrics

data_example

Format

A data.table with simulated data from a quasi-experiment (observational study) with five baseline covariates, a single timepoint binary treatment, and a continuous-valued outcome (metric for viewing time). There are 5000 unique units in the dataset, each in a single row, and 8 columns, described below.

id: Numeric ID of the observed unit. No repeated measures.
num_devices: The number of viewing devices recorded for the given user. A possible segmentation covariate.
is_p2plus: TODO: FILL IN. A possible segmentation covariate.
is_newmarket: A binary numeric indicator of whether the user falls in a region corresponding to a new market.
baselin_ltv: TODO:
baseline_viewing: TODO:
treatment: A binary numeric indicator of whether the unit received (non-randomly) the intervention of interest.
outcome_viewing: A continuous-valued measurement of viewing hours, the outcome of interest. Note that this mimics a metric derived from the viewing time, not the time itself.