Regression Analysis #2 (6 points) This part is based upon a birthweight dataset that we analyzed in class. Specifically, we have data on 1387 births. The following linear probability model (LPM) regression is used in order to explain whether or not a mother smokes during pregnancy (smoke = 1 if yes, 0 if no), where the explanatory variables are family income (faminc, measured in $1000's), mother's education (motheduc, measured in years of schooling), and race (white = 1 if mother is white, 0 otherwise). regr smoke faminc motheduc white, robust Linear regression Number of obs = 1387 F( 3, 1383) = 35 . 35 Prob > F 0 . 0000 R-squared 0 . 0640 Root MSE . 34864 Robust smoke Coef Std. Err. t P>It| [95% Conf. Interval] faminc - . 0022096 0005314 -4.16 0 . 000 - . 003252 - . 0011672 motheduc - . 0281384 0042072 -6. 69 0 . 000 -. 0363916 - . 0198852 white 0443496 0244681 1. 81 0 . 070 - . 0036491 . 0923483 cons . 54 62237 . 0559525 9.76 0 . 000 . 4364627 . 6559847 (a) (1 point) Interpret the motheduc slope estimate. (b) (1 point) In above regression, do we assume heterogeneity? 7 (c) (1 point) Explain why this LPM regression is not well-suited for predicting the conditional smoking probability (that is, Prob(smoke = 1 | x)) for a mother that has a really high faminc value (e.g., something like $1,000,000). (in one sentence)