Breaking Down A Contingency Table With SPSS



Breaking Down A Contingency Table With SPSS

[pic]

When you have a contingency table with more than two rows and/or more than two columns, you may want to follow a significant omnibus Chi-square with a set of analyses on 2 x 2 tables. For example, consider the relationship between gender and level of English using the Howell data file.

[pic]

As you can see in the contingency table, female students are more likely to be in the college prep English class than are male students, and male students more likely to be in general or remedial English than are female students. The Pearson (2 for this 2 x 3 table is 8.181, p = .017.

How should we break this table down? We could break it down into three 2 x 2 tables this way:

1. Gender x (College Prep, General) – that is, Gender x Engl ignoring remedial.

2. Gender x (College Prep, Remedial) – ignoring general English classes.

3. Gender x (General, Remedial) – ignoring college prep English classes.

To conduct the first analysis, just go to the data sheet, variables tab, and declare value 3 (remedial) of variable Engl as missing.

[pic]

Now repeat the analysis.

[pic]

SPSS reports (2 for this 2 x 2 table is 7.835, p = .005. Now you could go back and change the missing values specification for Engl to include remedial and exclude general to get the second analysis, and so on.

Instead of messing with missing values, you could use SELECT CASES IF to restrict the analysis to the desired columns:

[pic]

You could also select on the basis of both rows and columns (but not with these data), for example:

[pic]

Of course, it might well make more sense to break down the omnibus table by collapsing categories in various ways rather than excluding categories. For example, I want to compare the genders on proportion in college prep classes. I recode Engl into a new variable (CollPrep) where college prep (1) is recoded to 0 and general (2) and remedial (3) are collapsed into one group with value 1. Then I analyze the Gender x CollPrep table.

[pic]

[pic]

[pic]

As you can see, female students were significantly more likely than male students to be in college prep English, odds ratio = 5.54. I could then recode Engl into another variable where 0 = general and 1 = not general and then compare the genders’ enrollments in general English, and so on.

Return to Wuensch’s SPSS Lessons Page

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download