POLS5153-Writing Assignment 3

POLS 5153

Writing
Assignment Three

Please answer the following question in 6-8 pages of double spaced
normal font with inch margins. Assignment is due no later than November 14th,
2012.

Question
1

For this assignment I am interested to see if you can put all of
your regression knowledge together in a nal multiple regression paper. First
select a data set that will allow you to explore all of the assumptions held in
linear regression models. At this point lets select a dataset relevant to
political science and approach this as a political science paper. Within this
analysis please include some aspects of data transformation in your analysis.
Write this up and present it in a standard manner.

In this write up do the usual
things you have to this point, i.e. estimate a model and discuss the model. Be
sure to specify your equation, state your null hypothesis, the outcome of
hypothesis tests, the ability of the model to explain variance in Y and and ultimately
what the results of the analysis tell us, i.e. the substantive signi cance of
the your equation. This should also include a justi cation for the nal model
that includes an F-test. Be sure to present the estimates in an acceptable
table.

Next explore the regression
assumptions and how well your model meets these assumptions. Be sure to examine
the analysis for non-normally distributed error terms and the shape of error
variances as well as for the presence of multi-collinearity and outliers. If
you have data that uctuates over time be sure to test for the presence of
autocorrelation. If you nd any of these problems re-estimate the model with
some form of data transformations-i.e. weighted least squares, logged dependent
variable, Cochrane-Orcutt regression, robust standard errors etc… If in one
of your earlier papers you inadvertently selected time variant data it is
acceptable to re-estimate such a model and test for autocorrelation to
demonstrate your ability to estimate serially correlated error terms. Of course
you would also need to estimate a Cochrane-Orcutt model as well. Write up all
of your analysis and turn it in.