Your assignment is to use reliability analysis to create a multi-item index composed of three or more indicators. Then produce the appropriate univariate statistics for your index indicating central tendency and dispersion. Finally use your new index in a couple of crosstabs with other variables. While an index can in principle be constructed for either an independent or dependent variable, for this assignment, it is preferable to create an index for a dependent variable.

Begin by looking through the available codebooks for several ( >3; <6) questions that appear to probe a single underlying concept, attitude, characteristic, or other variable that interests you.

1. Describe in 25 words or less the concept that you want to measure. Also state the data set from which the questions are drawn.
2. List each question (indicator) you think could be part of an index of your concept. Use both the variable name (like CPSA5) and a brief (10-15 word) description for each.
3. Produce frequency distributions for the variables you have chosen. After defining missing values, use recodes to make all of the variables consistent in direction and range. For example, if you work with indicators of political interest make sure that they are all coded so that greater interest receives a higher value and all are coded on a consistent range, e.g., from 1—4 or 0—1. You need not submit these frequency runs with your assignment. Instead create a table like the one below to summarize your work. It should include the name of each candidate item, any values you declare as missing, the original range of values, the new range of values and the new label for the highest value. In the example below, variable Q50A “Should US and Canada have a Common Currency?” from CRIC 2003 has been recoded so that 1- “A very good idea” is the highest value and 4- “Not at all a good idea” is the lowest value.

4. Examine how well the items you are working with will fit together into an index using “reliability analysis” using either the reliability routine or the syntax editor in Webstats (SPSS). If you use the point and click method, be sure to select all of the descriptive statistics and summaries of means and variances as well as inter-item correlations.  The statistics subcommand will suffice using the syntax.
5. Eliminate any items that substantially weaken your index. Cronbach’s alpha and “alpha if item deleted” scores will be of particular use in deciding which variables should remain and which should go. Note also that an adjusted alpha substantially different from the simple alpha likely indicates scaling problems that must be addressed through recoding one or more of your original variables.  Systematically repeating this procedure, you will be able to determine which questions best make up an index of your concept or variable.
6. Report your best alpha score, noting whether it is acceptable (>.60) or not. If after several attempts you do not achieve (or very nearly approach) an acceptable alpha score, you should consider consulting with a TA.
7. Combine your selected variables to create a new variable using the compute command as outlined in the data lab. Using the frequencies program, calculate appropriate measures of central tendency and dispersion for your index. What does this tell us about your index? Based on these measures (as well as α), how would you characterize the variable in terms of its usefulness for further analysis?
8. Recode your index into fewer categories (e.g. into low, moderate, and high support) so as to make it easier to use in a crosstabulation.  Take care to ensure that the categories are meaningful, yet have sufficient respondents in each category.
9. Choose two independent variables you believe affect your index and construct hypotheses for them. Focusing on column percentages (i.e., compare you column percentages across the rows to identify the trends in the relationship) and the value of measures of association, write a short (4 or 5 sentences) paragraph describing each contingency table and whether the evidence supports your hypothesis. Briefly state which of the two independent variables best explains variation in your index and why you think that this is so.
10. Submit your answers along with the portions of output summarizing the final reliability run, two frequency runs index (unrecorded and recoded) and the two crosstabulations.

Assignments