BIO4180 Lecture Notes - Lecture 11: Overfitting, Analysis Of Variance
14 Dec 2019
School
Department
Course
Professor
Document Summary
If all variables are uncorrelated than simple and partial regressions will be the same. Both linear, so partial and simple regressions will differ. Because of the way that x1 and x2 are related to each other, the partial regressions relationship is negative. Difference between controlling for a variable and estimating y on x1 vs. ignoring it. If there"s correlation between the two, the regressions will be different since they are estimating different things. Bt=b0 + b1 * water temp (x1) B2= bt/g (mass, how much body temperature changes per unit change of body mass) Same data, scaled data differently, slopes are different. Hard to measure strength of effects when they"re measured on different scales. R2 = ssmodel/sstotal = sst-ssres/sstotal = 1-ssres/sstotal. Overall model is highly significant, high r2 value. But individually the yhat equation will deem each insignificant (x1 or x2 individually) since the other factor is involved. R2 : x1= x2 + x3 + x4.
