2024年4月12日发(作者：bgmbgmbgm欧美xx)

Practical Meta-Analysis

Effect Size Calculation Exercise

Calculate a standardized mean difference effect size for each of the following

examples using the Web-based Effect Size Calculator found at:

/resources/effect_size_

Note that this program calculates the standardized mean difference, the

correlation, and the odds-ratio and risk-ratio effect sizes. This exercise focuses

solely on the first of these.

Exercise 1: A study reports the mean self-esteem score on a standardized scale

(higher values indicating more self-esteem) for the treatment group of n = 25 as

127.8 and for the comparison group of n = 30 as 132.3. The standard deviations

are 10.4 and 9.8, respectively.

What is the ES for this data?

Exercise 2: A study reports a t-value of 1.68, favoring the treatment group. The

treatment group has 10 respondents and the comparison group has 12.

What is the ES for this data?

Exercise 3: A study reports that the treatment and control group sample sizes of

10 and 12 (df = 20). The study fails to report the means, standard deviations, or t-

value. However, it does report that the t-test was statistically significant (p = .

037).

What is the ES for this data?

Exercise 4: A study reports a correlation between the group variable (treatment =

1 and control = 0) and the dependent variable as r = .27.

What is the ES for this data?

ES Calculation Exercise – July 22-23, 2005Page 1Practical Meta-Analysis

Exercise 5: A study reports the following data:

Reading Achievement Test Scores by Group

Group

Learning Disabled Students

Low Achievers, Not Learning Disabled Students

Average Achievement Students

Advance Placement Students

Mean

55.38

59.40

75.14

88.00

n

13

18

37

22

The F-ratio for this data is reported as F

(3,86)

= 7.05. As the meta-analyst, you are

only interested in the difference between the first two groups. Unfortunately, the

study authors did not report the standard deviations in the written report.

What is the ES for this data?

Exercise 6: A study of adult psychotherapy reports the following data on the

percent of clients in the treatment and control conditions that are rated as

“improved” at post treatment. There were 42 patients in the treatment group and

29 in the control group.

Not Improved

Treatment

Control

What is the ES for this data?

Exercise 7: A study reports the frequency distribution for a variable of interest for

the treatment and control conditions (shown below) but does not report the means

and standard deviations.

Degree of Depression

No depression

Mildly depressed

Depressed

Profoundly depressed

What is the ES for this data?

Condition

Control

20

20

10

10

Treatment

15

15

15

15

32%

37%

Improved

68%

63%

ES Calculation Exercise – July 22-23, 2005Page 1Practical Meta-Analysis

Exercise 8 (extra credit!): A study using a nonequivalent comparison group

design compares partner reported domestic violence (as reported on the CTS/2

scale; this generates continuous type data for several forms of abuse) for men

court mandated to batterer’s intervention program to men sentenced to probation

only. The study measures a host of baseline measures (such as employment

status, marital status, age, etc.) that are know to be related to domestic violence

and includes these measures in a regression analysis predicting post-treatment

domestic abuse. The regression model also includes a dummy variable for

whether the man was sentenced to the intervention or probation only. You wish to

base the effect size on the “adjusted” estimate of the treatment effect from the

regression analysis. Information available includes the unstandardized regression

coefficient (-.523), standard deviation for the dependent variable, severity of

physical abuse (9.23), and the sample sizes for the intervention and probation only

groups (125 and 254, respectively).

What is the covariate adjusted ES for this data?

ES Calculation Exercise – July 22-23, 2005Page 1Practical Meta-Analysis