In the past, if a user tried to pass something impossible (e.g. pairmatch(pr ~ cost, controls = 100, data = nuclearplants)), they'd get the following:
## Error in pairmatch.matrix(m, controls = controls, data = mfd,
## remove.unmatchables = remove.unmatchables, : not enough
## controls in some subclasses
(Taken from some old slides that I'm in the process of updating.)
Now, the same code produces:
## Error in fullmatch.matrix(x = x, min.controls = controls,
## max.controls = controls, : negative 'omit.fraction' with
## 'min.controls' >= 2 not permitted
The former was much more clear to end-users. I get they say the same thing, but a user using e.g. pairmatch may not have any idea what omit.fraction or min.controls are.
I haven't looked into whether we've change the actual error, or (more likely) the order of operations has changed such that this second error gets hit earlier. Either way, I think we should try and catch this issue and error more informatively. Something like "The restrictions on the matched set sizes through controls/min.control/max.controls/omit.fraction are impossible to meet."
In the past, if a user tried to pass something impossible (e.g.
pairmatch(pr ~ cost, controls = 100, data = nuclearplants)), they'd get the following:(Taken from some old slides that I'm in the process of updating.)
Now, the same code produces:
The former was much more clear to end-users. I get they say the same thing, but a user using e.g.
pairmatchmay not have any idea whatomit.fractionormin.controlsare.I haven't looked into whether we've change the actual error, or (more likely) the order of operations has changed such that this second error gets hit earlier. Either way, I think we should try and catch this issue and error more informatively. Something like "The restrictions on the matched set sizes through
controls/min.control/max.controls/omit.fractionare impossible to meet."