201 Lab 5
Selecting Variables for a scale through reliability analysis

 ____________________________________________________________________________________________

Part I  

 As in the first lab, start up  SPSS and start a new data file that you  add data to for this lab.    Create participant code numbers for the first 10 cells (variable 1)  and label each of the next 5 variables with Q1, Q2, Q3, Q4, & Q5.

For this lab we will make use of this mini-questionnaire called the psyc 201 sociability questionnaire. 
As such we are trying to consider how sociable people are with this scale.

Add two more questions that you think may represent social styles of behaviour before you collect some answers from people in your group and others outside on campus or in other groups.

Next,  fill in the data set for 10 to 15  participants with their answers to 5 (+2) questions from the questionnaire.

We will then examine some functions for changing variable names and creating new variables from old ones using the Transform function. 

Next we will examine the relationships among the variables collected along with the new variable that we created using reliability analysis.


Changing Variable Names
You can switch between data and variable views by clicking on the tab at the bottom left of your SPSS data window. 

When inside the variable view you can click on the names and change them.  At this point change the names from Q1---Q7 to names that are meaningful to you (that briefly summarize the variables).

You can also adjust the number of decimals, variable type, etc. Try this out with your first data set, (later you may repeat this in  part II and then perform a reliability analysis with another data file).


Transformation

Once you have your data in a standard format

Add a new variable called "total" and  then perform a transform  where you compute the new variable by adding Q1 to Q5.  Save your data file.


Reliability Analysis

Click on Analyze then Scale then Reliability Analysis

Add each of the original variables (Q1-Q5)  (except for total and the two new questions) to the items box
   
(highlight then click the arrow to move them to the left)

Check to see if the model tab is on alpha and then click on the statistics button and check the following boxes: 
descriptives for: click scale, scale if item deleted;
 
then under

summaries
: click means and variances;
 then under
inter-item: click correlations
 then click continue to close the statistics window.

Click OK and let the analysis run
    (this will take a few seconds and the output window will open in front of you)

Scroll down the output window and look over the tables of statistics,
 in particular the alpha, corrected item-total correlation, squared multiple correlation, and the alpha if item deleted.
 

                    (ask for help if you cannot read the analysis results.
                         Note:  you may need to read the section in the book on reliability
                                   and Cronbach's alpha to make sense of this data)

Once this analysis has run and you have looked over the results try it again adding all seven variables
(include the two new ones) and run it again. Alternatively, add these two new variables and take out some of the first variables. Play around with the variables that you add in and take out to see what happens to alpha.

Create a word document where you answer the following questions and copy and paste your results from the  output relevant to each question to your word doc that you will place in the drop box for Lab 5.  

      What were the specific questions that you added (Q6 & Q7)?
    What can you say about the reliability of the total scale ?
    To raise internal reliability what items would you change or delete?
    Which item would you use as the best predictor of the total score?
             Why?

    Which items would make a good sub-scale if placed together?
             Why?


  
(consider these first for the data set of 5 questions then report the results from the second set with 7 questions )


  Part II - Optional

Go to data website and download the following data file
to your local area (201 folder).

INTDAT.xls       

Repeat Part I one using this data files. This set involves scores from a range of people on their score for a variable called Integration.  What this means is that those scoring higher are expected to maintain both their traditional cultural ways as well as "host" or mainstream cultural practices. 

open the first data file using SPSS and repeat part one transformation and reliability analysis with this data.
(Note when opening this file make sure the the box for "read variable names" is NOT checked, as there are no names in this excel file. Also you may need to open it first with excel and then save it as 97-2003 workbook. -as in Lab2)