Dietary wisdom
Due to renovations we like to simply summarize our week with this amazing poem by Robert J. Wherry Sr., famous for his extensive contributions to statistical models for prediction and factor analysis as well as to Industrial and Organizational Psychology.
Note the date of the source. Having a tremendously larger amount of computational power at out disposal compared to back in 1975, notorious overfitting seems easier than ever before. Thankfully, not only do remedies exist but also solutions for spotting backtest overfitting.
Models are fine and statistics are dandy
But don’t choose too quickly just because they’re handy
Stick to a model that’s been through the mill
Don’t try something new just for the thrill
A new shiny model is full of allure
But making it work is no sinecure
The more complex the merrier does not follow
The voluminous output may be hard to swallow
Too many variables and too few cases
Is too much like duelling at ten pace
What’s fit may be error rather than trend
And shrinkage will get you in the end
Know what you’re doing and do it well
Replictable findings are easy to sell
Be willing to progress one step at a time
A counterfeit dollar’s worth less than a dime
Now that I’ve warned you I’m ready to stop
And let you get back to tending the shop
