PA 705
Fall 2004

Univariate Statistics and Indexes

(updated Oct. 13)

Due Oct. 20

I. Please do Homework problem 6 of chapter 11  (p. 351, except you need not do the part of subsection "d" that asks for a percentile) and problem 3 from chapter 10 (p. 310) of ORB

Please show all your work!

II. Lab work: using the 2004 SF Citizen survey

You can find this in the lab on my computer/data on neptune/Naff, or on the Class web page

Begin by Going into Edit/Options. Make sure the General tab says to record syntax in journal and append. Make sure the draft viewer tab also has Display commands in log checked. Note that in the General tab you can also choose whether to display variable names or labels.

Please turn in all of your output, along with your analyses

A)  Choose 2 of the attitudinal (i.e., ask for respondents' evaluation) variables and obtain a frequency distribution (Use

From main menu go to Analyze/Descriptive Statistics/Frequencies

1. Graph these data in a way that best displays the results (From menu: Graphs)
    
2. Obtain the arithmetic mean, median, mode, standard deviation (From menu: Analyze/Descriptive Statistics/Descriptives.
    
3. Choose one of the demographic variables and create a graph that displays it.
    
4. Write a brief memo to the mayor analyzing your results and what it means for the City.  That is, explain what you found for the attitudinal questions, and the demography of the respondents.

B) Select several of the attitudinal questions that you think measure the same construct.

1. Run a factor analysis.  (From menu: analyze/data reduction/factor).  Choose Extraction--principle axis factoring. Choose rotation/verimax. Choose options/sorted by size.
    
2. Look at the rotated factor matrix.  It shows the correlation between the variables and each factor. You are looking at which variables load most highly on each factor.

   Give each factor construct a name that describes those variables.
    

3. Select one of the constructs with high factor loadings and make a note of which items load on it. 
    
4. Run a reliability test that includes those variables. (From menu Analyze/scales/reliability analysis).  Check the option under Statistics that will provide the "scale if item deleted." 
    
5. Create a new variable to create an index representing one of the constructs. It should represent the mean of the items (Use Transform/Compute/New variable).
    
6. Choose a suitable independent (demographic) variable, and compare the means and standard deviations (use Analyze/compare means/means) of the index for those demographic categories (e.g., for men and women, for owners and renters). 
    
7. Write a report to the mayor that describes the results of your factor analysis, the results of your reliability analysis, and the results of your comparison of the mean.