From various published research, I came to know that in cross-temporal meta analysis various studies conducted (during a specified time period) on a given population using a given measure are analyzed for ascertaining change in any phenomenon or variable over time. The mean score on the given measure of various studies are correlated with the time period of the study and the obtained correlation suggest the magnitude and direction of change in the given variable over a long period of time. I also came to know that while computing correlation between year of study and the mean score of the give variable, weighting is also done in a manner that methodologically more strong studies get higher weight. Further, some effect size measures (e.g. w) are also computed.
However, I am not having the clear idea of the computational steps involved in doing cross-temporal meta analysis as well as the methodological variants of it. The Google search yielded research papers that have used the said method and thus the computational details were presented with the assumption that readers are aware of the give methodology.
I would be much obliged if the esteemed members who have used this method or having some idea of it may extend some help in understanding this method. Any article or a ink to it dealing with the computational steps of the said method would be highly appreciated.
May I request to also share information about some software or SPSS macro/scripts/custom dialog for such analysis.
With warm regards
Rakesh Pandey
Hello Rakesh,
I guess you want to better understand how to perform meta-analysis from studies conducted or published in different years. If I am correct, then you need a computer software that will combine the individual effect sizes of interest (e.g. mean, odds ration, risk ration, etc.) and other attributes (sample size, standard deviation, confidence interval, p-value, z-score, etc) that you collected from the various studies. The computer software will combine them into one single OVERALL effect size and relevant overall attributes.
1) You may use STATA . The following reference will answer clearly your question:
Meta-Analysis: An Updated Collection from the Stata Journal Paperback – Mar 18 2009
by Jonathan Sterne (Author)
2) You may also use CMA (Comprehensive Meta-Analysis), which I most familiar with. Here is the link where you get a 10-day free trial:
http://www.meta-analysis.com/pages/demo.php
http://www.meta-analysis.com/pages/book_introduction_to_meta-analysis.php
3) Using SPSS to run meta-analysis seems not straightforward - relatively. Read John Reece's comment in a discussion I initiated on LINKED-IN following the link below:
https://www.linkedin.com/grp/post/1879554-5956197072154669060
I hope the above helps.
Joseph
Dear Joeseph
Thanks for your helpful reply.
I am pasting bellow a link to an article that uses cross-temporal meta analysis. It will help you and other esteemed members to understand about the statistical method that I want to know.
https://www.researchgate.net/profile/Brad_Bushman/publication/5342670_Egos_inflating_over_time_a_cross-temporal_meta-analysis_of_the_Narcissistic_Personality_Inventory/links/0912f5134dae28cafc000000.pdf
Thanking you once again
Regards
Article Egos Inflating Over Time: A Cross-Temporal Meta-Analysis of ...
Thank you for sharing the articles, which clarify the specific context of your question to me.I suggest you follow the commands under the meta-regression section in STATA or the follows the steps in CMA software to do the cross-temporal meta-analysis (CTM). In SPSS you may use the regression command.
It transpires from the different articles that CTM is more like a form of multivariate analysis involving on main dependent variable -the effect size from each study- and one main independent variable (IV) which is the year of publication -or time.
For example, one of the article ran a bivariate correlation between the mean of the outcome of interest and the year published. Another article performed a mixed effect model, which is a form of META-REGRESSION that includes other moderating variables in addition to the main independent variable (year of publication).
Knowing your dependent variable (i.e. effect size of interest) and your IV (year of publication) you can follow the regression command in SPSS. For instance, to perform the hierachical linear regression models the following link is easy to follow:
https://www.google.ca/webhp?sourceid=chrome-instant&ion=1&espv=2&ie=UTF-8#q=hierarchical%20multiple%20regression%20using%20spss
Please let me know your thoughts on the above.
Joseph
Dear Joeseph
Thanks for your reply. I know how to perform the hierarchical linear regression models in SPSS. What is not clear to me from the papers that I attached is -
1.how they weighted the mean score in relation to sample size
2. While doing regression whether the said studies used year as predictor and the mean sore of the given variable as criterion. Or they used some regression equation to predict the mean score and used the predicted mean scores and DV.
If you have got the idea how the CTM is done then may I request you to detail the steps so that I can do the needed analysis using SPSS. What is most important to me is to know how the data for IV and DV are extracted from past studies. Whether it is just noting the year of publication and entering it in one column (in SPSS) as IV and then entering the mean score (reported in given study) in the second column as DV. I think that DV score used in the said studies are not just the mean score reported in the published paper rather it is some score derived from the published mean score.
Thanking you once again for your help
Regards
- Badges
- Science topic