[Maarten Buis]

About the predicted values, I remember seeing regression plots containing predicted values beyond 0 and 1 in some R tutorials.

My suspicion is that they used the poisson family instead of the binomial family. This would work, and combined with robust standard errors would make as much sense (or as little, depending on your preferences) as a linear probabilty model.

The funny thing is that the glm command with log link works for upto 3 IVs. (I find it unfair, but a robot is a robot after all)

The number of independent variables is irrelevant, all that matters is whether you get predicted values larger than 1.

[Sonnen Blume]

My suspicion is that they used the poisson family instead of the binomial family. This would work, and combined with robust standard errors would make as much sense (or as little, depending on your preferences) as a linear probabilty model.



The number of independent variables is irrelevant, all that matters is whether you get predicted values larger than 1.

Thanks Maarten! This is really an important lesson to learn.