Quant formulas

[MrSmart]

Do we have to memorize the formula for the estimated variance for the prediction error? Seems too long.

 

[onlysimon]

first order predictor = SEE
second order predictor = [SEE^2 (1+1/n)]^0.5
after that I’ll be picking A, B or C as I think is likely to be closest.

 

[MrSmart]

What?

 

[onlysimon]

you are talking about..
sf^2 = SEE^2 * [ 1 + 1/n + (x-xhat)^2 / (n-1)sx^2 ]
.. the last term is only significant at the 3rd decimal place, so pointless to memorise it.

 

[MrSmart]

Hmm, need to see that for myself.
Thanks for the tip.