Can anyone show me how to meta analyse data from the following studies? Our analysis is a little different to traditional meta analyses in that the studies we are analysing aren't interventions & aren't assessing relationships between variables. Instead, at a single time point, they assess distress levels among participants with a particular cosmetic condition.
These studies report Means &SDs based on various distress measures. Our aim is to get a pooled mean and SD (across those studies using the same specific distress measure).
MetaXL and other softwares seems to require effect sizes which arent reported.
Thanks!
Thanks I'm finding learning R difficult - its user interface seems trickier than SPSS / excel. I also only have M SD and Ns from these studies not confidence intervals or effect sizes.
Other options for its use are found in other free statistical programs similar to SPSS such as JASP or JAMOVI, which include the R packages for metanalysis and is more accessible to those who do not handle the programming.
Dear Glen Jankowski,
You have asked the following question which reads as follows: Can anyone show me how to meta analyse various Means &SDs from cross sectional, single group studies? Please deifne first what you mean by: meta analyse various Means &SDs from cross sectional, single group as it will make it easier to understand exactly what it means and to finally be able to give a correct or a logical answer to the question.
Best Regards
Desalegn Abraha
Hi Glen,
Depending on the study design(s), there are various free online effect size calculators available, or you could even compute yourself in Excel, if by 'single group' studies you are referring to within-subject designs? Please see relevant article link for details on effect size calculation - https://www.frontiersin.org/articles/10.3389/fpsyg.2013.00863/full
Andy Field's SPSS/R textbooks also state how to calculate various effects sizes (e.g. Pearson's r, Cohen's d, partial eta squared etc.) manually, I believe, so may be worth having a read?
Once you have your effect sizes across all studies, you may then be able to proceed with meta analysis using R, the below guidance may help with this:
https://bookdown.org/MathiasHarrer/Doing_Meta_Analysis_in_R/
https://bookdown.org/MathiasHarrer/Doing_Meta_Analysis_in_R/pool.html
Hope this helps and good luck!
Best,
Jason
Thanks Desalegn Abraha Gebrekidan I've clarified the question above.
Thanks Cristian Ramos-Vera - I'll look into those if SPSS and Excel aren't able to perform this.
Thanks Jason Round really appreciate it. I will try and find an automatic effect size calculator, as doing it manually intimidates me. I'm reluctant to use R as I've haven't learnt this programme before especially using script so would prefer excel or spss. Do you know of an effect size formulae for excel based on Ms and SDs only? These studies are cross sectional (there is no pre or post scores, just scores at one point). I've tried to clarify the type above.
Hi Glen Jankowski
Thanks for clarifying the study design for me, and my apologies, I did not see the full question you posted before I replied.
To the best of my understanding, an effect size is usually used to quantify the magnitude of a group difference (e.g. Cohen's d) and/or relationship (e.g. Pearson's r) found, so typically pre-made excel calculators are designed for that purpose, e.g. Cohen's d calculator for between- and within-subject t-tests.
However, after reading fully, I'm not entirely sure what the most appropriate effect size would be based on your original question/aim. This online calculator allows you to compute various effect sizes and provides more detailed information on each calculator - https://www.psychometrica.de/effect_size.html (calculator 13 allows for pooled SD calculation from up to 10 different group means). Is there any existing previous work that has conducted the specific form of meta analysis you are referring to? If so, I would read/ask how they approached this to know what software, effect size, etc. may be most appropriate and replicate if possible.
Best,
Jason
Dear Glen Jankowski,
Thank you very much for your clarification of the questions above as per your request.
Best Regards
Desalegn Abraha
Dear Glen Jankowski,
As you know, the most important step (as well as first step) in a Secondary research, typically Systematic Review (SR) is defining "Review Question" (RQ). It's seem that your RQ is a relational RQ. Therefore, when relational SR / MA targeted the cross-sectional studies, you should be extracted the appropriate "Effect Size" (ES) data (or relevant raw data).
Although cross-sectional studies seem to be single-group, they are classified into at least two groups based on the dependent variable (or outcome) and usually the study hypothesis evaluates based on comparing the exposure (independent variable) in the outcome groups (usually 2 groups).
If your independent variable is quantitative, you should be extracted the 3 data items for each group; the specific sample size, mean and standard deviation (sd). Therefore, you totally should be collected 6 data items (n1, m1, s1, n2, m2, s2) for calculating appropriate ES that named; Mean Difference (MD) and Standardized Mean Difference (SMD). Of course as you know the last ES is named "Cohen's d", too.
Best regards
Thanks Jason Round very helpful response and it was my fault not making the original question clearer. That helps to know about the purpose of effect sizes (I thought I was missing something),. There is actually a similar study and the authors have shared their excel file with us very kindly.
Abbas Keshtkar thanks for your help too. That is helpful but it doesn't fit our research question or the studies we are reviewing - any subgroup analyses are peripheral and they don't include any independant variables consistently.
- Badges
- Science topic
- Similar topics
- Biological Science
- Histology
- Sectioning