Final Assignment - Winter, 2000
Due Date: I will not be in Regina after April 20, so would prefer that all the work for this class be handed in by April 20. If you cannot get the work completed by April 20, but have it completed by noon, April 25, give it to Alison Hayford (CL 220) and she will arrange to get the work to me. If you anticipate that all the work for the class will not be completed by April 25, arrange to discuss with me the possibility of an incomplete.
Background to the Assignment. This final assignment asks you to use various statistical procedures to analyze some of the attitude or opinion variables in the Survey of Student Attitudes and Experiences, Fall 1998. The data from this survey are in the 703.sav data file. More specifically, I would like you to analyze the opinions in the “Social and Political Views” section on pages 4 and 5 of the questionnaire. However, you may also want to include other opinion variables, from other sections of the questionnaire, in the analysis.
You will probably not be able to find very powerful explanations of any of the variables. When this survey was designed, there were no specific hypotheses that guided these questions – rather, they were developed from the interests of the students and instructor. In addition, more possible explanatory variables should have been included, but again since the aim of the survey was to develop a profile of students, not a statistical or theoretical model, little attention was paid to this issue. But there are enough variables in the survey that you should be able to obtain some statistically significant results, even if the overall model is not very satisfactory.
For each variable used in the statistical analysis, make sure that it is appropriately coded or constructed. For example, some of the variables such as income or grade point average should be recoded to ensure that they have interval level scales. Other variables will have to be constructed by using the recode or compute procedure.
For each of the questions, hand in or describe what you did – in enough detail so I could reproduce the results. For each question, state any assumptions and describe the findings – what results are statistically significant, what your hypotheses are, what assumptions might be violated, etc. Make sure you provide a short description of the results in words, so that a non-statistician would know what the results demonstrate. As much as possible, attempt to provide a sociological interpretation of the results, as well as a statistical explanation.
1. Initial regressions. Take two of the variables V1 to V9 and regress them on a number of independent variables in the data set. For example, you might regress V4 on sex, age, importance of religious values, etc. or V5 on income, support for the various political parties, and having a dependent. Describe the results. (Note that you are later asked to compare these to the results in part 5, so you may want to modify these after completing some of the other parts of this assignment).
2. Factor Analysis. Conduct at least two factor analyses using the variables V1-V9 and other opinion variables you might have an interest in (e.g. you could use the variables in question 17 and some of the multiculturalism variables). One of the factor analyses should include just the variables V1-V9 and the other should include V1-V9 and some additional opinion variables. It may not matter much which extraction method you use, since the different methods tend to give the same result. However, use the varimax rotation method, and under options request that the factor scores be sorted by size (this will provide for easier interpretation). Under descriptives, request that the correlation coefficients be provided. After you try various factor analyses, decide on two that appear to be reasonably interpretable, provide the main statistics for these, and write a short description of the results.
3. Construct Two New Variables. From one of the factor analyses, construct two new variables representing the sum of the values of the variables in two different factors. For example, suppose V1, V5, and V8 are associated with one of the factors in the factor analysis. Use the compute procedure to construct a new variable which is the sum of these three variables. Where the variables have factor loadings that differ in sign, you will have to recode the variable with the different sign so that the direction in which the two variables are connected to the factor has the same direction for each. Obtain frequency distributions for each of the two variables constructed.
4. Analysis of Variance. Conduct two one-way analyses of variance for each of the two new variables as dependent variables. As independent variables, use variables that you hypothesize will have an effect on the new variables as dependent variables. Describe the patterns of means, and the anova table.
5. Multiple Regression. Obtain regression equations for each of the two new variables as dependent variables. Attempt to find independent variables that are reasonable in light of your general sociological knowledge and that explain the variation in the dependent variables. Describe the results in words and, where possible, compare with the equations in part 1.
6. Conclusion. Write a summary conclusion concerning your findings. From what you find, make suggestions concerning how a survey of this sort might be changed – e.g. other variables to add, revision of existing variables, changes in wording of questions, sampling procedures.