Quote:
Originally Posted by █████
The modeled data has the same shape as the model, but the model is too steep, and reaches the top, which it shouldn't, according to the data it models.
I'm afraid I haven't explained the figure well enough. The model isn't trying to match the data as tightly as possible, which is why they don't match. :-)
Each player has some true value P, which we would like to model.
We observe data, D, which is a function of P.
The data, D, has extra variance since for many players we have a small number of samples (extra variance makes the shape less steep).
We make a model, M, that approximates P by subtracting out the extra variance from D.
Does that make sense?
Last edited by Agthorr; 01-23-2009 at 10:42 AM.
Reason: Fixed grammatical error