F Statistic Suggests Some Sort of Functional Form Problem at 5 % Level.

Now, F statistic suggests some sort of functional form problem at 5 % level.

6. (i) Acceptability for federal funded lunch program is highly related to weak economic conditions. Also, it seems realistic that if a percentage of students eligible for the federally funded school lunch program is controlled for, than expected value of percentage of students living in poverty does not depend on expenditures per student and size of enrollment. So, lncprg is a good proxy for poverty.

(ii) It seems that in first case (column (1)) we work with omitted variable bias. As suggests column(2) bias is positive. This also makes sense:

1. Naturally that, on average, poverty and success at the school is positively related (coef. near lncprg in the column(2) is an evidence for this.)

2. Also, on average, in poor areas expenditures per student are lower than in rich.

So, excluding proxy for poverty from error term decreases effect of spending per student which remains statistically significant at 5% level in the column (2) due to t-statistic (7.75/3.04 > 2).

(iii)Yes, negative and statistically significant coef. (at the 5% level) with t – statistic of |-1.2|/0.58 > 2 near log(enroll) suggests that ceteris paribus, on average, schools with bigger enrollment has lower pass rate .

(iv)On average, ceteris paribus, increase in percentage of students eligible to federal funded lunch program by 1 percentage point leads to 0.324 percentage point fall in pass rate. (recall that lncprg and math10 are measured in percentages).

(v)Model with lncprg explains about 19% of variation in pass rate and about 3% without lncprg. So, the most of the variation in math10 we control for is explained by lncprg.

7. (i)

(ii)

8.