Oh, and the paper's title is "The Robust Beauty of Improper Linear Models in Decision Making" and, in my brief skimming, had little to do with marital happiness although there is mention of that at one point.
The correlation of rate of lovemaking minus rate
of arguments with these ratings of marital happi-
ness was .40 ( p < . 0 5 ) ; neither variable alone
was significant.
Which is why it can be important to include interactions to your linear models, which would most probably have resolved this. This also allows better quantification of individual and joint effects than handcrafting composite variables (as far as your sample size etc allows it).
I think the baseball stat OPS (on base percentage plus slugging) fits this mold of "improper linear model" quite well. There are much more robust measures of hitting now, but OPS works better than it ought.
Comments
Should add "(1979)" to the title.
Oh, and the paper's title is "The Robust Beauty of Improper Linear Models in Decision Making" and, in my brief skimming, had little to do with marital happiness although there is mention of that at one point.
Which is why it can be important to include interactions to your linear models, which would most probably have resolved this. This also allows better quantification of individual and joint effects than handcrafting composite variables (as far as your sample size etc allows it).
In this example, what would "including interactions" look like?
I think the baseball stat OPS (on base percentage plus slugging) fits this mold of "improper linear model" quite well. There are much more robust measures of hitting now, but OPS works better than it ought.