Page 94 -

80 CHAPTER 4 Statistical analysis

4.4.4 TWO-TAILED T TESTS AND ONE-TAILED T TESTS
In some empirical studies, the hypothesis indicates the direction of the difference.
For example, you may expect the use of word-prediction software to improve typing
speed. In this case, the hypothesis of the study will be:
Individuals who use word-prediction software can type faster than those who do
not use word-prediction software.
How does this hypothesis differ from the original hypothesis? In the original
hypothesis, the direction of the difference is not specified, implying that the use of
word-prediction software may improve typing speed, reduce typing speed, or have
no impact on typing speed. In the hypothesis specified in this section, we expect
the use of the word-prediction software to either improve typing speed, or have no
impact at all. In this case, a “one-tailed t test” is appropriate. A t value that is larger
than the 90% confidence interval suggests that the null hypothesis is false and that
the difference between the two means is significant.

4.5 ANALYSIS OF VARIANCE
ANOVA is a widely used statistical method to compare the means of two or more
groups. When there are only two means to be compared, the calculation of ANOVA
is simplified to t tests. ANOVA tests normally return a value called the omnibus F.
Therefore, ANOVA tests are also called “F tests.”

4.5.1 ONE-WAY ANOVA
One-way ANOVA is appropriate for empirical studies that adopt a between-group
design and investigate only one independent variable with three or more conditions.
Let us revisit the word-prediction software study from Section 4.4.
Suppose you are also interested in a speech-based data-entry method and would
like to compare three conditions: text entry using standard word-processing software,
text entry using word-prediction software, and text entry using speech-based dicta-
tion software. The independent variable of the study has three conditions. With a
between-group design, you need to recruit three groups of participants and have each
group complete the text entry task using one of the three methods.
The data layout for running one-way ANOVA using SPSS is similar to that for
the independent-samples t test. Table 4.6 presents a data set for the one-way ANOVA
test. The Coding column marks the group that each data point belongs to. Normally
we use 0 to mark the control group (those who used the basic word-processing soft-
ware); 1 and 2 are used to mark the group who used the word-prediction software and
the group who used the speech-based dictation software. When using SPSS, only the
third and the fourth columns need to be entered.