Psych. 280: Psychological Statistics/Methods
Fall 2005
SPSS Instructions for Independent samples t tests
Independent Samples t Test
You will begin your analyses of the depth-of-processing experiment data by conducting three
independent samples t tests (one for each possible pairwise comparison across the three levels of the
independent variable: structural vs. semantic, structural vs. self-referent, and semantic vs. self-referent).
Note that you will NOT actually be reporting these t tests in your APA-style paper for the project; when
an IV has three or more levels, the appropriate analysis to run is an ANOVA, which you will learn about
next week.
We have already entered the data from the experiment and have saved it to Blackboard in the
“Assignments” folder, under Project 3. The file is called project 3 data.sav. The first several cases of the
file appear as follows:
Notice that the file contains a variable called “condit” (for “condition”) in which the numbers 1, 2, and 3
are used to represent the three levels of the IV (1 = structural, 2 = semantic, 3 = self-referent).
Remember that you can always toggle to “variable view” and click on “value labels” to view the labels
assigned to each number. The file also contains a column for each of the following four DVs: recallc (#
of words correctly recalled), recallf (# of words falsely recalled), recogc (# of words correctly
recognized), and recogf (# of words falsely recognized).
To conduct an independent samples t test, go up to the “Analyze” menu and click on “compare means”
and then “independent samples t test.” Click on the DV you want to analyze (you can do all four at once,
if you like) and click on the arrow to put it in the “test variable(s)” box. Click on your IV and put it in
the “grouping variables” box. You should see the following window:
Notice the question marks next to condit in the Grouping Variable box. Click on “Define Groups” to tell
SPSS what numbers are used to represent each of the two groups (e.g., 1 and 2 for structural vs.
semantic). Once you’ve done so, click Continue and then OK. SPSS will bump you to the Output
window where you can examine the results of the t test.
Below is sample output from an independent samples t test (NOT from our class experiment):
Notice that SPSS computes the means and standard deviations for the two levels of your independent
variable. Before you examine the results of the actual t test, you should check the results for Levene’s
Test for Equality of Variances, which assesses whether the variances of the two groups are significantly
different from each other (i.e., whether the homogeneity of variance assumption has been violated). If
the p value for the Levene’s test (in the “Sig.” column) is greater than .05, then the variances are not
significantly different from one another (i.e., the homogeneity of variance assumption has been
satisfied), and you may use the t value and degrees of freedom in the row marked “equal variances
assumed”. If the significance value for Levene’s Test is less than .05, you will instead need to use the
values reported in the row labeled “equal variances not assumed”. Note that the observed t value is the
same in either case; the bottom row just uses adjusted df.
In the output shown here, the variances are not significantly different, so we may use the t test results
from the row labeled “equal variances assumed.” We see that the observed t, with 18 df, is -2.478, and
the p value is .023. Since p < .05, this test is statistically significant. We’d now look at the means to
determine which group had the higher mean. It looks like the attractive job candidates received
significantly higher evaluations than did the unattractive job candidates. When you’re writing these
results up, be sure to include the t (rounded to two decimal places), df, p value, and the mean and SD for
each group. See p. 286 of your text for how to write up the results of an independent samples t test.
(However, note that your book underlines M, SD, t, and p, whereas you should italicize them).
Important details for writing a Results section involving an independent samples t test:





You should italicize M, SD, t, and p.
If p = .002, you can’t say that p < .002. Instead, say p < .003 or p < .01.
When p > .05, report that the difference was NONsignificant, not INsignificant.
Correct way to report nonsignificant differences: t(31) = .38, n.s.
Round everything to 2 decimal places.
Tree-saving hint: SPSS likes to put each output run on a separate page. To conserve paper, when you
choose “Print” from the “File” menu, change “General” (in the top left) to “Layout” and choose to print
two pages to a page.
Practice Exercise for Independent-samples t test
(You don’t have to turn this in; it’s just to give you practice.)
Let’s say I’m interested in the social skills of people who grew up with siblings and people who are only
children. I might hypothesize that people who had siblings are more socially skilled than only children
because they had lots of practice negotiating with brothers and/or sisters.
H0: s = o (mean social skills for kids with siblings is equal to mean social skills for only children)
H1: s > o (mean social skills for kids with siblings is greater than mean social skills for only
children)
Note that this is a one-tailed hypothesis. What would a two-tailed one look like?
Let’s say I have people answer a questionnaire in which they report whether or not they had siblings (1=
yes, 2 = no) and then rate their level of social skills on a 1-9 scale, with higher numbers meaning “more
socially skilled.” (Obviously, there are problems with having people report their own level of social
skills, but this is just a hypothetical situation…) Imagine the data are as follows:
SIBS
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
SOCSKILL
8
6
3
9
7
8
4
3
9
7
2
4
3
8
1
4
7
2
5
1
Go ahead and put these data into SPSS and conduct an independent samples t test as practice. What do
you get? Note that SPSS conducts a two-tailed test by default. If you want to make sure the t value is
significant for a one-tailed test, you need to look up the critical value in the t table, and make sure that
the observed t (from the SPSS output) exceeds it. (Note, though, that unless your mean is in the opposite
direction of what you predicted, your t will always be significant at the one-tailed level if it was
significant at the two-tailed level. Why is this?) Should we reject the null hypothesis? What should we
conclude?