Comparison of group means: t-statistic

Question 1

What are t-tests tell us?

Answer 1

• Trying to find differences
• Only one outcome measure
• Interval/Ratio Data

Question 2

t-statistics compare 2 ____

Answer 2

means

Question 3

These tests are ____ statistics

Answer 3

• parametric

Parametric statistics:
- Homogeneity of variance
- Random Assignment
- Normal Curve
- Ratio/Interval

Question 4

The Concept of Testing Mean Differences

Answer 4

• Need to know the differences between means (degree of separation between groups) and variation (variability within groups)
• In theory, if the experimental treatment was effective, and all other factors are equal and constant, all subjects in each group would have the same score, but subjects in each group would have different scores.
• In other words, there would be between group differences, but no variance within the groups
• Thus, all differences could be explained by the effect of the treatment
• Reality doesn’t work this way due to personal characteristics, measurement error, and other extraneous variables

Question 5

Draw a visual of what two different means would look like with bell curves

Answer 5

Left Graph: Difference; no overlap in bell curve
Right Graph: Groups not different; large overlap

Question 6

What is a way to make variation smaller in a within subject (same people) design?

Answer 6

• Take repeasted measures. This keeps make variation smaller and increase power
• This does require a wash out period

Question 7

If Levens test is small, what should we do?

Answer 7

• Breaks homogeneity of variance. Need to look at additional information for a conservative test
• Smaller than 0.05 would indicate being homogenous

Question 8

Determine the degrees of freedom for 20 subjects split into two groups

Answer 8

• df: n-1 (Must be done for both groups)
• df = 18

Question 9

Why do we want a high t score?

Answer 9

• High negative or positive is good. Explains difference in means

Question 10

If the data violate homogeneity, what do you do?

Answer 10

Need to look at the bottom line of the middle table in SPSS as it is automatically adjusted for

Question 11

One tailed vs two tailed power

Answer 11

• One tailes is more powerful

Question 12

When sample sizes are unequal, what can happen?

Answer 12

• Larger variance, large sample t-test becomes less powerful.

Question 13

Paired t-test

Answer 13

• Repeated measures on same subjects
• Analyzes the different scores (d) within each pair so subjects are compares only with their match

Question 14

When putting paired data into SPSS it must be placed ____ the table

Answer 14

ACROSS

Question 15

t-tests are used to compare ____ means only!

Answer 15

2

Question 16

The more comparison one makes, the more likely to make a ____.

Answer 16

Type I error (finding a difference when none exists)

Question 17

If you use an alpha of .05, we have 5% error for a single test. With multiple tests, cumulative error ____

Answer 17

Greater than .05

Question 18

What else is an important measure to calculate for t-tests?

Answer 18

• Effect Size = Change in Score/Avg SD
• Greater than 0.8 is Large Tx Effect!

Question 19

Independent t-test data is place ____

Answer 19

• Down!
• Should only be group number and one score