Title: Testing Hypotheses About 2 Related MeansThe PairedSamples t Test
1Testing Hypotheses About 2 Related MeansThe
Paired-Samples t Test
2Hypothesis Testing Terminology
- Null hypothesis no difference
- New treatment produces SAME results
- College grads work same as pop 40 hr/wk
- Alternative hypothesis null is false
- New treatment changes the cure rate
- College grads do no work 40 hr/wk
3Rejecting the Null
- If the observed sig. level is small, lt.05, you
REJECT the NULL - When you reject the null, the confidence interval
does not include 0 - You can also test at the .01 level, and/or do a
99 CI
4Paired t Test
- For when you have 2 measurements, or observations
under 2 conditions, or studies with pairs of
subjects or measurements matched in some way - Testing the null that the average difference b/w
a pair is 0 - Chapter example endorph.sav
5Why Use a Paired Test?
- Makes it easier to detect true differences when
they exist - Do not have to stick to before/after, can compare
wife/husband, siblings, etc. - Do not use without a true pair
- Evaluating difference the sign tells you the
direction of the change
6First Step Check Data
DIFF Stem-and-Leaf Plot Frequency Stem
Leaf 1.00 0 . 3 4.00 1
. 0127 5.00 2 . 00458 1.00
3 . 0 Stem width 10.00 Each leaf
1 case(s)
7Second Step Run Test
- You can use a one-sample t test if you just use
the difference - If you have before/after scores, then use the
paired-samples t test - Your mean difference, t value and 2-tailed
significance will all be the same! - Usually easier just to use the paired-sample t
test
8Paired-Sample t Test
Analyze?CompareMeans?Paired-Samples T Test
9Third Step Make a Decision
Based on the 2-tailed significance level, we
REJECT the NULL hypothesis of no difference NULL
also stated mean difference 0 two means are
equal The sig. is lt.0005, but, COULD WE BE WRONG?
10For APA-Style Write-ups
- Paired-Samples t Test
- Green Lesson 22
Green also discusses effect size indices d
for one-sample d or ?2 for paired-samples