You are interested in estimating the relationship between school-level performance on a fourth-grade math test (% passing) to socioeconomic characterisitcs of students attending school. The variable free, measured at the school level, is the percentage of students eligible for the federal free lunch program. The variable medinc is median income in the ZIP code, and petsgle is the percent of students not living with two parents in the ZIP code. lexppp is the log of expenditure per pupil. Regression A: math4 = 96.77 - 0.833 * petsgle n = 299, R2 = .380 (6)
(1.60) (0.071)
Regression B: math4 = 93.00 -0.275 * petsgle - 0.402 * free n = 299, R2 = .459
(1.63) (0.117) (0.070)
Regression C: math4 = 24.49 -0.274 * petsgle – 0.422 * free - 0.752 * Imedinc + 9.01 * lexppp n = 299, R2 = .472
(59.24) (0.161) (0.071) (5.358) (4.04)
Regression D: math4 = 17.52 - 0.259 * petsgle - 0.420 * free + 8.80 * lexppp n = 299, R2 = .472
(32.25) (0.117) (0.070) (3.76)
If you had to choose one of the four regressions as your best estimate of the effect of petsgle and obtain a 95% confidence interval, which would you choose? Why?