1. The effects of scaling data on OLS statistics
Suppose a data set of 1388 observations (N = 1388) was analyzed using OLS to examine the factors that influence the birth weight of children. The regression results are as follows, with standard errors in parentheses:
bwght^ = 116.97 (1.05) − 0.46 (0.09) cigs + 0.93 (0.03) faminc
where
bwght = birth weight, in ounces
cigs = number of cigarettes smoked by the mother while pregnant, per day
faminc = annual family income, in thousands of dollars
R
2
= 0.021
SSR = (10,000)
SER = (1,000)
Suppose you would like to measure the dependent variable bwght in grams, rather than ounces. Let bwghtg be the birth weight measured in grams.
Hint: There are approximately 28.35 grams per ounce.
Without rerunning the regression, you know that after transforming your regression to change the units of the dependent variable, your new slope parameter β1^ would be 28.35. You also know that the t statistics as well as the R-squared, for the regression with bwght as the dependent variable, are those of the regression with bwghtg as the dependent variable.
Leave a Reply