% turnout in 2000 election 55.155 48.755 Mean
Difference = 6.395 % turnout in 2000 election Equal variances assumed 5.647 0.021 2.938 49 0.005 6.4 2.179 2.022 10.779 Equal variances not assumed 4.315 35.942 0 6.4 1.483 3.392 9.409

**The T-Test in
Research**
**The T-TEST in
SPSS **

**Independent**
samples**Paired** Samples
(Matched Measures) -- for samples of matched
cases

_{1} :
States outside the south had a higher voting turnout
in 2000 than states in the "old south."_{0 }:
States in the old south had a turnout that was
*higher or equal* *to* turnout in
non-southern states._{1} is a directional hypothesis, and the
data are in the direction that might **disprove**
the **null** hypothesis, H_{0}.

**Group Statistics**
**Former confederate state, 1=yes,
0=no**
**N**
**Mean**
**Std. Deviation**
**Std. Error Mean**

**Independent Samples Test**
**Levene's Test for Equality of
Variances**
**t-test for Equality of
Means**
**F**
**Sig.**
**t**
**df**
**Sig. (2-tailed)**
**Mean Difference**
**Std. Error Difference**
**95% Confidence Interval of the
Difference**
*Lower*
*Upper*

**T-Test** of voting turnout that the difference is
significant at .01 for a **two-tailed** test, but what
about a **one-tailed** test?

**halve** the 2-tailed probability to
get a one-tailed probability level

**both** tails;

**smaller** *absolut*e critical value of t,
for example, 1.65 is less than 1.96**plus or minus**
3.03 has .0015 in each tail of the distribution

**one-tailed **test, we are interested
only in the .005 associated with the **critical
value**, 3.03**The important thing
to note** is that a one-tailed test makes it "easier"
to reject your null hypothesis by requiring a smaller
critical value to reject, and, by implication, tests
statistics calculated for a two-tailed test are
**twice** as significant (i.e., the associated
probabilities are divided by two and therefore 1/2 the
size)**proportions** rather than **means**?

**mean** of a **dichotomized**
variable, measured 0 for one category, and 1 for the
other*proportion* of those in
category 1 (or *percent*, if multiplied by 100)
.

**variance** of a dichotomous variable when p =
proportion in one category and q = proportion in
the other is simply, **variance =
pq****standard
deviation **is the square root of **pq **for
dichotomous proportions**standard
deviation** becomes square root of **pq**
divided by N**T-Test** to test for differences in
**proportions**, as well as in **means**.

Go here for an example using the
**Paired Sample T-Test.**