I have made several submissions and gotten what I perceive to be very puzzling results.  Do you have to make your submissions in the same order as the sample?  I figured that if you entered a submission and had the patient ID number in the first row, claims truncated in the 2nd row and your prediction in the 3rd row, it didn't matter what order your IDs were in.  If you have to make your submissions in the same order as the sample my results begin to make sense to me.  It could be that I am hopelessly lost, but after very discouraging initial results, I have vastly simplified my model, and I really can't imagine it is less predictive than a constant value.  If it is, c'est la guerre...

Yes it matters. Make it in the same order as the sample.

On the one hand, whew, maybe I'm not as dumb as I was beginning to think I was, on the other hand, I suppose it is obvious since I asked the question that I missed it,but does it say that anywhere? I'm sure it probably does, but I was unable to find it. I'm sure a bolded "Submissions should be given the same order as the sample" note would be helpful to people other than me. First place, here I come! If this answer has helped you out a lot, you should consider hiring me because I am unemployed.

It's posted some where on this forum.

This is not something I would even have thought of, so thanks for pointing it out! It should definitely be in an FAQ somewhere.

I don't mean to be rude, but what kind of IT company are we dealing with here? Having to submit in the same order as the provided CSV is mystifying. Not that it's a problem, but it's an unnecessary query that I have to write.

Oh and I'd love to employ you Doc, but I'm an only an employee myself - and probably in a different country to you :)

 

I am not surprised that order is important. I do not expect a smart computer program to calculate the score for the submission and with stupid program that does not care to check that the member id in the first column are in the same order, in order to change the order, it is obvious that order is important. I think that it could be better simply to ask people to submit only one column of expected days and in this case it could be obvious that order is important and the files could be smaller(so it could also save time and space for the site). I suspect that the software that calculates score does not look at the columns of memberid and I suspect that replacing the numbers of the first 2 columns with 0 can also lead to the same result with a smaller file but I did not try it.