Several SPSS procedures test for statistically significant differences in means for an interval-level dependent variable when analyzed by two or more groups classified on a nominal scale. All of these procedures conduct an analysis of variance and produce an F-ratio as the test statistic.

Like z-scores and t-scores, F-ratios are interpreted with reference to a table of critical values--such as the F table on pages 710-717 in Kirk. Note that the magnitude of the F ratio must be interpreted with reference to two degrees of freedom. One for the numerator in the ratio and another for the denominator.

ANOVA is the general procedure for analysis of variance in SPSS, but ANOVA involves ideas not covered in our course. Instead, we will feature two very similar procedures:

• ONEWAY is the basic procedure for analysis of variance for interval-level variables when analyzed by one independent nominal variable.
• MEANS was developed just to compare means for an interval-level variable when broken-down into categories of several levels of nominal variables. It also generates an analysis of variance and produces the useful eta² statistic for explained variance. I prefere MEANS over ONEWAY, but you should know both.

The syntax for the ONEWAY and MEANS commands are given below:

ONEWAY

"varlist" names the dependent interval-level variables to be analyzed. More than one dependent variable can be analyzed on a given ONEWAY run. Suppose that you wanted to analyze the percent of land under cultivation (in the POLITY data the variable would be landcult) across nations in the major regions of the world.

"varname" names the single nominal-level variable used to classify the several groups being compared. For example, "Region" in the table below might be the variable for the groups.

(min,max) indicates the range of values that embrace the groups being compared. For the "Region" variable in the POLITY dataset, specify (1,6).

REGION seven world regions (POLITY dataset) Valid Cum Value Label Value Frequency Percent Percent Percent ANGLO-AMERICAN 1 5 4.5 4.5 4.5 EUROPEAN 2 17 15.3 15.3 19.8 LATIN AMERICAN 3 23 20.7 20.7 40.5 ASIA 4 15 13.5 13.5 54.1 MIDDLE EAST 5 11 9.9 9.9 64.0 AFRICA 6 32 28.8 28.8 92.8 EASTERN EUROPE 7 8 7.2 7.2 100.0 ----- ------ ------ Total 111 100.0 100.0

ONEWAY
varlist BY
varname
(min,max) .

ONEWAY output

oneway landcult by region (1,7).

MEANS

 

"varlist" names the dependent interval-level variables to be analyzed. More than one dependent variable can be analyzed on a given MEANS run.

"varname" names the multiple nominal-level variables used to classify the several groups being compared. For example, "Region" in the table below might be the first nominal variable for the groups and "country" might be the second.

The same variables analyzed under ONEWAY are used in this analysis with MEANS:

2 means landcult by region /statistics=anova.

MEANS
varlist BY
varname BY
varname ...
/STATISTICS
=ANOVA.

MEANS

output