Never-treated individuals

[johnnybonney]

Allow option in ES() to use never-treated individuals as a control group.

[davidnov]

I have added this as an option to ES().

I will add a description of this option to the front page, but the gist is that the argument is never_treat_action and it takes on values in c('none', 'exclude', 'keep', 'only'). It is meant to work with the case where, by assumption, units missing onset_time_var are considered to be never treated.

A bit more detail:

• none -- (default) no action is to be taken, implying that there is no never-treated group. If it turns out that there exist units in long_data with missing onset_time_var, using never_treat_action = 'none' will throw an informative error and ES() will break.

• exclude -- exclude any observations with missing onset_time_var.

• keep -- keep units with missing onset_time_var and include them together with the not-yet treated when forming control groups.

• only -- keep units with missing onset_time_var and only use them (i.e., exclude the not-yet treated) when forming control groups.

For never_treat_action %in% c('exclude', 'keep', 'only'), if there are no units with missing onset_time_var, ES() will throw an informative error, suggest never_treat_action = 'none', and break.