Field Experiment

Is it possible to randomize exposure to the different levels of the predictor?

Is there natural variation in the predictor in the field?

Quasi-Experiment

Is the variation over time in the predictor exogenous?

Selected endogeneity threat

Sensitivity Approach

Is it (still) likely that an unobserved confound would alter the conclusion?

Is a strong and valid instrument for the predictor available?

Are instrument-free approach requirements fulfilled?

For example, are the sample size and the distribution of variables adequate?

Is it possible to triangulate the results with a controlled study?