SPSS DATA ANALYSIS

Lesson 2: Examining Your Output We just ran crosstabs to test a hypothesis with two variables, one nominal (SEX-independent variable) and one ordinal (AFFRMACT-dependent variable).  As you can see, the categories of the independent variable are found across the top in the columns and the dependent variable information is found down the side forming the rows.  Each square is known as a cell and within each cell is the frequency (or count) and the column percentage.  You can also find the row totals and column totals, which are sometimes referred to as marginal. In our example we know the following is true:  7.8% of men in this sample strongly support affirmative action policies, whereas 11.4% of women do;  We can also look at grouping at a glance and concede that 15% of men and 18.5% of women support affirmative action policies in comparison to 84.9% of men and 81.5% of women oppose these policies.    The bottom right cell in the table is where we can find that we had 1,904 people answer this question as our sample.   Lesson 3: Interpreting Crosstabs Researchers run crosstabs to determine whether there is an association between two variables.  Also, crosstabs may tell us other important things about the relationship between the two variables, including the strength of association, and sometimes the direction of the association.  FYI, the direction can only be found when both of the variables in your table are greater than nominal. Ask yourself the following questions after you populate your crosstabs: 1. Is there an association between the two variables? If you answer yes to this (or maybe), then move to question 2. 2. What is the strength of association between the two variables? If BOTH variables are ordinal than move to question 3. 3. What is the direction of association? Is there an association? What we are trying to determine here is whether knowing the value of one variable will help us predict the value of another variable.  In other words, if gender is associated with preference to affirmative action policies.  In order to find out is SEX and AFFRMACT are associated, we read across the rows of the dependent variable to see if there are differences in the column percentages.  When you look at the data for each category, you see that men and women are approximately the same in their preference. Let’s see if there is an association between one’s gender and one’s preference for affirmative action policies.  We will want to first read across the rows of the DV to see if there are differences in the column percentages.  In reading across the first row, we see that women (11%) are more likely than men (8%) to strongly support affirmative action policies.  4  The crosstabs suggest that women are a little more likely than men to support affirmative action policies, but knowing the respondent’s gender does not help us enough in determining whether they support affirmative action policies. How Strong is the association? There are tests for this; however, there are a few rules you can apply here to see the rudimentary version with this data.  One rule is that the larger the percentage differences across the categories, the stronger the association.  On the contrary, the smaller the percentage differences across categories, the weaker the association between the variables.  There is also a rough 10-percentage-point rule which states that is the percentage point difference is 10 or more, the relationship between variables is probably worth examining further. Tips for Identifying Independent and Dependent Variables Identifying the independent and dependent variables are tricky sometimes.  Here are some “hints” to try and help you master this skill: 1. Restate the hypothesis as an “if ___[independent variable], then _____ [dependent variable]” statement. Ex: If you are a female (IV- gender), then you are more likely to support affirmative action policies. 2. The independent variable is going to influence the dependent variable. Ex: Ask yourself if support for affirmative action can influence gender.  Obviously you can reason that this is incorrect and will switch the two to make a more sensible statement.  Can gender influence support for affirmative action?  Because you can answer “maybe” than you know that gender is the IV and support is the DV. 3. The independent variable comes first in time “before” the dependent variable. 4. Ex: Ask yourself which came first, gender or support?  Since gender must be determined before support for affirmative action policies in our example, than we are sure that gender is our IV. Conclusion When looking at crosstabs researchers move from searching for the answer to what and looking for why.  Social researchers will make the shift from examination of relationships to explanation.   Follow-up Questions 1. What is bivariate analysis?  2. What are three general questions to ask when trying to understand a crosstab with two ordinal variables?