This handout states what I would like you to do regarding the term project for BUEC 333. You will get a fair project grade provided that you fulfill all the requirements stated below. You will be penalized only for a bad interpretation of the results, not for bad statistical results.
The deadline for submission of the project is Thursday of the last week of classes. There are absolutely no extensions.
The project will comprise of a 5 - 7 pages (single-spaced) report (please type) on your findings according to the format stated in the Project Outline. (Include computer output as an appendix and do not count this as part of the 5-7 pages.) The computer output will be discussed in tutorials, so be there. You may find the following references helpful:
| 1. Multiple regression | Ch. 13 |
| 2. Model Building | Section 14.1 |
| 3. Specification Bias | Section 14.5 |
| 4. t-test and misc. | Section 12.8 |
| 5. F-test | Section 13.7 |
| 6. Standard Assumptions | Section 13.3 |
| 7. Multicollinearity | Section 14.6 |
| 8. Heteroskedasticity
- LM test |
Section 14.7 |
| 9. Autocorrelation
- Durbin-Watson test |
Section 14.8 |
State your topic and the independent variables chosen. Mention the period of study and the level of significance chosen for the tests of hypotheses. In no more than three sentences, analyze your audience (i.e. who are they? why do they care?). What economic theory could be used to suggest a model and variables? Does your model reflect economic theory?
For each variable, state (1) what the variable is, (2) what adjustment, conversion, or transformation has been done (e.g. seasonal adjustment, monthly to quarterly conversion, logarithmic transformation, etc), (3) what is the rationale of choosing the variable (for independent variables only), (4) what is the expected impact of the variable (for indep. vars. only), and (5) the source of the data series; in particular, report the CANSIM series number.
1. Report the estimated equation together with the (1) R-Squared, (2) R-Bar-Squared, (3) t-statistics (put underneath the estimated coefficients), and (4) Durbin-Watson Statistic (or Durbin-h, which ever is appropriate).
2. Interpret the Adjusted-R-Squared, and explain why it is preferred to the unadjusted R-Squared.
3. Interpret one (or more if you wish) of the estimated coefficients, and do a t-test on it.
4. Do an F-test for the overall significance of the regression.
1. Multicollinearity
a. Report the simple correlation coefficients and F-tests on subsets of the regressors. What are the implications of these for multicollinearity?
b. Explain what is multicollinearity and which OLS assumption is violated by its presence.
2. Autocorrelation
a. Do the Durbin-Watson test (or Durbin-h, if appropriate).
b. Explain what is autocorrelation and which OLS assumption is violated by its presence.
3. Heteroskedasticity
a. Report visual inspection of plots.
b. Do the LM test in the text.
c. Explain what is heteroskedasticity and which OLS assumption is violated by its presence.
Briefly present your principal findings. Suggest direction for further research. Briefly state the consequences of multicollinearity, autocorrelation and heteroskedasticity on the validity of hypothesis testing and statistical inference. Based on the findings in the statistical analysis above:
(i) What can you say about the tests of hypotheses performed in section C?
(ii) What can you say about the OLS estimates of your model (i.e. Is OLS BLUE? ) ? Is economic theory supported by your model? Do you think your model is misspecified?
The following is a rough guide as to how much space a student should devote to each section of the project report.
The basic project report can be written in 5-7 pages single space or 10-14 pages double space with a regular font size (ie 12 in MS word).
A: Introduction
- can be done in one paragraph and should not exceed more than 1/2 of a page!(single spaced)
B: Data definitions
- roughly 1 page(single spaced)
C: Statistical results and hypothesis testing
- roughly 1 page(single spaced)
D: Statistical analysis
- roughly 2-3 pages(single spaced)
Q. Some students do not know the deadline.
A. The deadline for the project is the last Wednesday of classes, absolutely no extensions permitted.
Q. For autocorrelation, one way to check is the visual inspection. However plots a and b in the book (checking autocorrelation) are hard to tell apart for rho =0 and rho=0.3. Is that a mistake or basically visual inspection is not that reliable after all?
A. Yes, visual inspection for autocorrelation is not very reliable.
There are more FAQ's on the FAQ web page.
There are a number of options available to retrieve CANSIM data. See the lab Web page for details.