This paper proposes and theoretically justifies bootstrap methods for regressions where some of the regressors are factors estimated from a large panel of data. We derive our results under the assumption that T/N→c, where 0≤c∞ (N  and T  are the cross-sectional and the time series...