Economic statistic lab

Ec170 Exam 3 Review Exercise

Part A. Graded exercises. Show your work to get credit for your answers. Use your own paper. This assignment is due at the beginning of the next class.

1. Lab 2 recap:

We found in the multicolinearity part, that multicollinearity might be a problem, even though the R2 value really wasn’t that high. The best thing to do when evaluating multicollinearity is to examine the correlation matrix for the defined problem. The rule of thumb (high R2 and few significant X variable coefficients) is useful, but the correlation matrix gives you more information.

In the heteroskedastcity part, the BP test showed no heteroscedasticity (be careful to watch out for pasting errors like I had in class!). The calculated F decision rule statistic was low (about 1.087), and the critical value can be estimated using a critical F with 3 numerator df and 40 denominator df (because most tables don’t have a row for 45).

In the serial correlation part of the lab, serial correlation WAS present, based upon the Runs Test. There were 3 runs, so the number of runs fell well outside the confidence interval of 27.667 ± 7.044. We should probably try to do something about it.

Assignment.

Address the serial correlation from Lab 2. In order to do that, let’s try to model the time series issue. Estimate the following three regressions, and choose the best one.

A) TOTCOMP=f(FRINPERC, time)

B) TOTCOMP=f(FRINPERC, lagged TOTCOMP)

C) TOTCOMP=f(FRINPERC, time, lagged TOTCOMP)

The “time” variable is already in the data set, but you will have to generate the lagged TOTCOMP variable. In order to do this, select and copy the TOTCOMP data column, then paste it into an empty column one row lower the other data – this will give you the TOTCOMP data for the previous period lined up with the other data.

When you run a regression with the lagged data, you will lose one observation (because there is no lagged value for the first observation). Make sure to select data rows that have values for all the relevant variables when running your regression.

2. Use the Z Data Set posted on Blackboard. Estimate this regression: Y = f(X1, X2, X3). Analyze the results in the usual way, including checking for the three problems we discussed. Revise the model (drop one or more X variable, transform the data, etc.) In order to develop a better model than the initial one. Explain in detail how you arrived at the new model.