Meta analysis

Question 1

7 steps of research for literature reviews

Answer 1

1: select research questions
2: selecting bibliographic databases
3: choosing search terms
4: applying practical screening criteria
5: applying methodological screening criteria
6: doing the review
7: synthesizing the results

Question 2

validity in meta analyses depends on

Answer 2

reviewer’s expertise
critical imagination and quality of available literature

Question 3

what step of research extends to numerical aggregation rather than having a qualitative focus

Answer 3

7 - synthesizing the results

Question 4

reasons for meta study

Answer 4

power
reliability increase
noise reduction
practical design issues in individual studies

Question 5

goals of meta analysis

Answer 5

estimate population effect
each study provides estimation so not sure about true effect
explain variabiity

Question 6

steps of MA

Answer 6

1: objectives description
2: inclusion/ exclusion criteria
3: literature search
4: developing standardized protocol to evaluate quality of studies retrieved
5: coding form
6: meta analytic procedures
- pooling of study results
- testing for homogeneity of effect sizes
- selecting moderator variables
7: describe results, conclusions, limitations

Question 7

how to deal with different effect sizes

Answer 7

translational scales for different measures of effect size

Question 8

statistical goals of meta analysis

Answer 8

estimate combined ES
estimate confidence interval around ES
- variability within and between studies (s^2 and t^2)
explain variability between studies (potential moderators)

Question 9

how do you combine effect sizes?

Answer 9

they dont have the same weight so dont average them
- instead you should give weight to larger studies and less weight to smaller ones as their evidence is less strong

Question 10

fixed effect model

Answer 10

assumes there is true effect in population and all error is sampling error

study weight: w= 1/s^2

Question 11

random effects model

Answer 11

study weight w=1/(s^2+t^2)

Question 12

are studies homogeneous

Answer 12

test with Q-statistic
- is the between study variance
if p value is smaller alpha, then heterogeneous
- but with few studies there is a power problem

Question 13

heterogeneity tests: some measures

Answer 13

t^2= estimated amount of heterogeneity (insensitive to number of studies and precision)

I^2= total heterogeneity
- not sensitive to number of studies
- not always adequate measure for heterogeneity
-suggested as criterion to go to subgroup or modertaro analysis

H^2= total variability, sampling variability
- ratio of SDof the estimated overall effect size from a random effects MA compared to SD of fixed effects MA

Question 14

what bias do you have to be aware of in MA

Answer 14

publication bias
- file drawer problem
- accessibility bias
- duplication
- country/ language/ culture

Question 15

statistical procedures against publication bias

Answer 15

• funnel plot by Eggers (plot effect sizes of studies against their precision)
• fail safe number, add z value of amount of studies that could be missing and see if still sig