Question 4 (20 pts): Let\'s try to predict Academy Awards (frequently known as t

Question

Question 4 (20 pts): Let's try to predict Academy Awards (frequently known as the Oscars) Best Picture nomination winners. For a sample of 105 films nominated for the best picture in the period from 1984 through 2004 we have the following information: whether or not the film won an Oscar for best picture (win-1 if a film won, 0 otherwise); film's total number of Oscar nominations (nominations); its number of Golden Globe wins (gglobes); and whether or not the film was a comedy (comedy-1 if comedy, -0 otherwise) You estimated the linear probability model for winning and got the following results: win =-0.355 + 0.049 nominations + 0.1 17gglobes-0.183comedy (0.097) (0.013) [0.078[0.013 (0.026) 0.027 (0.102) [0.089] 105, R2-0.4525 n usual standard errors are in (), heteroskedasticity-robust standard errors are in [.] (a) What is the baseline group in this regression?

Explanation / Answer

a. The baseline group in the regression is whether a movie is comedy or not. It is basically the value of dummy that is equal to 0.

b.
If nominations increase by 1 unit, then the probability to win improves by 0.049 units.
If gglobes increase by 1 unit, then the probability to win improves by 0.117 units.
If comedy increase by 1 unit, then the probability to win decreases by 0.183 units.

c.

Ho: Bi = 0
Ha: Bi not equal to 0.
where i is the coefficient of each variable

we construct a t-test statistic, t = (Bi - 0)/se(bi). se(bi) is standard error of bi.

Below are each test statistic when usual standard errors are used:

the critical value at 5% level = 1.98 (approx)

We notice that nominations and gglobe are significant but comedy is statistically insignificant at 5% level.

If we use heteroskedastic adjusted standard error then -

In such a case, all parameters are statistically significant at 5% level of significant.
For coefficient of comedy variable, it does matter which standard error is used.

d. We should use robust standard error. In context of Linear probability model, heteroskedasticity is a major problem. Hence it is recommended that we should heteroskedasticity adjusted standard errors.

Variable Coefficient t-statistic nominations 0.049 3.769231 gglobes 0.117 4.5 comedy -0.183 -1.79412

Related Questions

Navigate

• Browse (All)
• Browse Q
• Subjects

---

---

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.