10. Comparing Two Population Means

10.10 SPSS Lesson 7: Paired Sample t-Test

To follow along, load in the Data Set “Methadone.sav”:

SPSS screenshot © International Business Machines Corporation.

As set up, the file has two dependent variables. This “within subjects” dataset is fundamentally multivariate. When we did the paired t-test by hand we converted the multivariate data to univariate data by taking differences. SPSS will do the differences behind the scene and you won’t actually see them. Run the t-test by picking Analyze \rightarrow Compare Means \rightarrow Paired -Samples T-Test:

SPSS screenshot © International Business Machines Corporation.

Move the two variables into Pair 1 and hit OK (Options again allows you to specify a confidence intervals percentage):

SPSS screenshot © International Business Machines Corporation.

The output is:

SPSS screenshot © International Business Machines Corporation.

The first two tables are descriptive statistics. The last table gives the stuff we want: \overline{D} = 0.9615, s_{D} = 10.7067, the confidence interval

(10.11)   \begin{equation*} -5.5084< \mu_{D} < 7.4315, \end{equation*}

t_{\rm test} = 0.324, \nu = 12 and p = 0.002 for the two-tailed hypotheses pair

    \[H_{0}: & \mu_{D} = 0 \]

(10.12)   \begin{equation*} H_{1}: & \mu_{D} \neq 0. \end{equation*}

The very low p-value (0 in this case) and the absence of 0 in the confidence interval guide us to reject H_{0}, the differences are significantly different from zero.

The standardized effect size and strength of association for the paired t-test are

(10.13)   \begin{equation*} d = \frac{t}{\sqrt{n}} = \frac{\overline{D}}{s_{D}} \end{equation*}

and

(10.14)   \begin{equation*} \eta^{2} = \frac{t^{2}}{t^{2}+n-1} \end{equation*}

respectively.

License

Share This Book