Why are some weights bigger than 1 and some weights smaller than 1 when using the CEM command in Stata? How does CEM generate those weights for each control group? Can we use the weights to treat the data as weighted data and do the following regressions? When we do CEM using the cross-sectional data and then expand the matched sample into panel data, the cross-sectional data might be balanced, while the panel data might not be balanced between the treatment and control groups among various covariates. How would this influence the following estimation? And how can we deal with it? When doing CEM with panel data, we could match at each time point before treatment; in such a scenario, the control group might not be matched at every time point, so when we merge the matched sample into the panel structure, which control group should we keep?