We’ve arrived at the Final Project, which will force you to tie together everything we’ve done in the course as you present a full statistical analysis (with heavier emphasis on the material after completion of the second project). As with the other projects, I will grade this as if it’s been submitted as part of your dissertation in an effort to give you feedback as to what we’re looking for. You will need to find your own data set for this project, and import it into SPSS. Try to find one with many variables, some of which are categorical and some of which are quantitative. Begin your writeup by discussing your original impressions, that is, which variables you expect to work together to predict other variables, etc. It’s fine if your summary at the end completely deviates from your initial impressions, I’m just curious as to what you expect to prove with your analysis.
The first portion of the project will be a review of the material from earlier in the course. You will have to do pre-screening, as usual, to account for outliers and any type of irregularity (problems with normality, linearity, and homoscedasticity). Then you will run some form of ANOVA/MANOVA/ANCOVA/MANCOVA to determine group differences, that is, which of your independent variables significantly affects the values of the dependent variable(s). Next, you’ll need to use multiple regression to build a best fit line that best predicts the value of the dependent variable (either enter or forward method may be used). If you want to
practice path analysis, I’ll certainly encourage you to build a model and estimate the causal relations among the variables, but this has been a large source of consternation, and we already demonstrated this during the second project, so this is completely optional.
The second portion of the project will highlight the newer techniques that we have learned. First, we will employ factor analysis on the independent variables to determine if there are underlying components consisting of variables that are working together. Finally, you will have to use discriminant analysis and/or logistic regression to predict which group or value of the dependent variable we most expect linear combinations of the independent variables to fall into. You will then summarize your findings as if you had completed a full analysis of the data (it’s ok if some of your results did not work out as you originally hoped that they would, I’m more interested in seeing you try to work through these techniques).

1) Data Screening:
• Are there any missing data points?
• Are there any outliers? (compute the Mahalanobis distance for all variables, independent and dependent, and look up the cutoff for removal from the table on page 357).
• Create a scatterplot matrix of all IVs and the DV. Any ’boomerang’ shaped plots should be transformed with a natural log transformation.
• Run normality plots with tests and interpret Kolmogorov-Smirnov.
• Create and display a residual plot, and conclude whether there is linearity, normality, and homoscedasticity.
2) ANOVA/MANOVA/ANCOVA/MANCOVA: (you do not need to run all of these)
• Develop appropriate research questions/hypotheses for an ANOVA/MANOVA test
• Display a factor interaction graph and interpret potential interaction between variables.
• Run ANOVA/MANOVA, and display the full ANOVA/MANOVA summary table. Interpret the results, indicating which factors influence the dependent variable, and whether there is any factor interaction.
• Display a Bonferroni comparison table if appropriate when there are several groups. Determine which groups are different from one another.
3) Multiple Regression:
• Run multiple regression using the ’enter’ OR ’forward’ method. Display the best fit line as an equation, and identify which variables are significant (if you ran forward method, tell which variables were included/excluded, whereas if you ran enter method, all variables will be included, but some will be insignificant).
5) Factor Analysis:
• Run factor analysis in SPSS using at least EIGHT independent variables (it’s ok if you haven’t used eight up until this point).
• Determine the appropriate number of components (at least 2) to retain using eigenvalues, the variance, a scree plot, and residuals.
• Interpret and label components (note which variables are loaded onto each component, as well as the strength of each loading). Try to best name each component based on a common characteristic.
• Summarize the underlying structure that exists between your independent variables.
6) Discriminant Analysis: (either use this or logistic regression or both to predict group membership)
• Run discriminant analysis in SPSS using the stepwise method (you must choose a categorical variable as your DV).
• Indicate which independent variables show significant group differences as well as the number of functions that are significant.
• Calculate the effect size for the function(s).
• Display the function with unstandardized coefficients (make sure to check this box under statistics).
• Interpret the standardized coefficients (which variables were most influential, and whether this relationship is positive or negative).
• Discuss the cutoffs for membership in each group using the function(s) (this comes from the functions at group centroids box).
• Address the percentage of cases correctly classified for the original variables (classification results box).

Finish the project by summarizing what you have found. You are expected to report your findings in a format similar to the Results section of a research journal, as you will do in your disseration. The textbook provides examples of how to write up your results at the end of each chapter under ’Presentation of Results’. Discuss your overall findings (which variables work together, what your models say about the relationship between the variables).
I’m not looking for any specific result at the end of the project. You’ve worked hard: show me what you can say about the variables, and what direction you’d like to take this work if you had more time and/or experience. Thank you for a great semester, and I hope that you have some confidence that you can accurately conduct a statistical analysis as you work on your respective dissertations. My office (and email) is always open to you if you need advice.

We’ve arrived at the Final Project, which will force you to tie tog...