Preparing Your Data
1. Select the following ten variables for your analysis. You will need an interval-ratio dependent variable, and the following independent variables: two appropriately coded dichotomous nominal variable, two appropriately coded multi-category nominal variables, two ordinal variables, interval-ratio variables, and one index (a variable measuring a latent concept comprised of several existing variables).
2. Clean your Data
A. Recode any missing data
B. Reverse code variables and create dummy variables as necessary.
C. Check your coding with cross-tabulations (append the SPSS output)
3. Construct an additive index (scale) variable. It must be computed from at least three other similar variables (either nominal or ordinal, as long as they all have the same category scheme).
4. Data Transformations A. Use the ladder of powers to examine alternate transformations both interval-ratio variables. Make sure to use the following: x3, x2, identity,√ , ln(x) 1/√ , 1/x, 1/ x2, 1/x3.
B. Create a table of the transformations and discuss which transformations may be useful and why.
2
5. Interaction Term
A. Compute an interaction term from two of your existing variables to be included in one of the intermediate models in your analysis (not the final model).
II. Paper Sections
1. Hypotheses
A. Discuss your empirical expectations for the independent variables that you are interested in and formally state a hypothesis for each (you must have at least four).
2. Data Section
A. Discuss the specifics of the dataset, including the particular details of the year you are using (e.g., the size of the sample, the non-response rate, a description of the sampling frame, type of probability sample, how the data was collected, etc.).
B. Compute a table (Table 1) of the Kolmogorov-Smirnov tests for each transformation of both of the two interval-ratio variables. Make sure to include the test statistic and probability value of each transformation. Remember that you don’t need to actually include a transformed variable in your final analysis, but you do need to discuss the advantages and disadvantages of doing so (both theoretically and empirically), and how you would select an appropriate transformation.
C. Construct a table (Table 2) of the Variance Inflation Factors (VIF) for all of the independent interval-ratio and ordinal variables. You don’t need to compute the VIF for the nominal variables, but these variables still need to be estimated in the models you use to compute the r-squared value of the interval-ratio and ordinal variables. Interpret the table and diagnose any potential multi-collinearity issues.
D. Construct a table of descriptive statistics (Table 3) and use it to explain your variables (how they variables are coded in the analysis and what the descriptive statistics of each mean) to your reader in detail. Make sure to discuss the mean of the variables (or the percentage for the dummy variables).
E. Explain your variables and how they are currently coded (not the steps you took to recode) to your reader in detail.
F. Discuss how you dealt with missing data.
3. Methods
A. Discuss the assumptions of the OLS regression model and assess how well your model meets these assumptions.
3
B. Discuss your analytic strategy (what you are going to do in the following analysis and why).
C. Discuss the logic you are using to add variables to the analysis (you need to have at least four separate models).
4. Results Section.
Conduct a multivariate regression analysis. Compute at least four separate models (all shown on the same table), beginning with the primary variable of interest, and adding additional groups of variables in each model. At least three of the independent variables need to be significant in the final model.
A. Construct a correlation table (Table 4) of all of the variables in the model using Pearson’s r as the measure of association (since SPSS can’t compute polychoric correlations even though they would be the most appropriate to use here). Discuss the table and diagnose any potential multicollinearity problems.
B. Compute a table (Table 5) of your multivariate regression analysis. Make sure the table can stand-alone, and that it includes the r-squared valued and N at the bottom of each column.
C. Make sure to also include the interaction term in in one of the models (but not the final model) and discuss whether it is significant or not.
D. Discuss model fit using r-squared.
E. Interpret the significant coefficients in the final model.
F. Interpret any changes in significance or effect size as it relates to mediating/intervening variables.
G. Evaluate your hypotheses.
