12-10-2013, 08:16 PM
First of all, 123k IS huge. Good enough for the second largest in history by my measure anyway. Nevertheless, I am not getting my hopes up, nor am I going to even worry about it until the Bonnie counts get rolling in April (probably late April at that).
Secondly, I don't think that the correlation has been all that bad. Looking at the correlation between jack return and following year's adult return (going all the way back to when they starting counting jacks at LG in 1980) gives an R-squared of .81. Pretty good in my book. And the only instances I can see that the jack return would have severely under-predicted the ensuing adult return were the '09 Jack > '10 Adult, and the '11 Jack > '12 Adult. It is actually slightly more common for the jack return to over-estimate the adult return using that .81 R-squared (3 over-est outliers vs. 2 under-est outliers).
Luckily, our agencies use far more complex models however, than simple jack > adult regression analysis.
[signature]
Secondly, I don't think that the correlation has been all that bad. Looking at the correlation between jack return and following year's adult return (going all the way back to when they starting counting jacks at LG in 1980) gives an R-squared of .81. Pretty good in my book. And the only instances I can see that the jack return would have severely under-predicted the ensuing adult return were the '09 Jack > '10 Adult, and the '11 Jack > '12 Adult. It is actually slightly more common for the jack return to over-estimate the adult return using that .81 R-squared (3 over-est outliers vs. 2 under-est outliers).
Luckily, our agencies use far more complex models however, than simple jack > adult regression analysis.
[signature]