Meta-analysis

Question 1

Where is a meta-analysis used?

Answer 1

-JDM
-EWT
-Therapies
-Expert evidence
-Interventions like CI

Question 2

What is a meta-analysis? (7)

Answer 2

-Find patterns and relationships between factors (which characteristic biggest impact in JDM)
-Examine results of several prev studies in topic area
-Secondary data
-Conducted when a lot of research or if inconsistent
-Studies from diff time periods, cultures and locations
-Look at direction and size of of effects
-statistical analysis

Question 3

What must be done carefully when doing a meta-analysis?

Answer 3

-Considering what what studies to include
-Data should be collected in similar way and be comparable

Question 4

What is the 5 step process of a meta-analysis?

Answer 4

1. Research qu identified
2. Systematic review addressing qu to see what studies are relevant
3. Data extraction: summary of data from each study. Numerical or categorical measures. Stat analysis
4. Standardisation and weighting studies: DV measured diff across studies, need to eb comparable
5. Final estimate of effect

Question 5

What is the best evidence synthesis?

Answer 5

-Only include studies that have good methodology so best evidence is used
-systematic critical assessment of the quality and risk of bias
-unreliable studies can impact on conclusions

Question 6

How can they be used by expert witnesses?

Answer 6

-Arvey
-Used meta analysis to prove that data that expert witness was using in an age discrimination case was an outlier

Question 7

What are the strengths of meta-analysis?(6)

Answer 7

-Robust data: best evidence. Published, peer reviewed. Recalc so know sound.
-Generalisable: larger sample.
-Reliability: patterns and trends. Where data agrees/doesn’t
-No ethical issues: 2ndary data. Don’t collect from ppt
-Objectivity: numbers BUT not indepth
-Useful for lots of factors: where future research should be directed

Question 8

What are the weaknesses of meta-analysis? (7)

Answer 8

-small studies: JDM shadow and mock, have own S+W. Affects data, conf v’s
-experimenter bias: select the studies they use. could just agree.
-publication bias:only successful research published. Howells only exception
-risk of unpublished: not peer reviewed, no checks for errors
-sample issues
-objective: what but not why, no depth.
-only main effects: miss factors for further research