Hi,
Maybe you should consider an ANCOVA model using the no-nitrogen yield as the baseline and the applied nitrogen level as the co-variate. As each study has repeated yield measurements with different nitrogen levels, you may add a random intercept and slope in your model; or alternatively use GEE to account for the covariance of the measurements. Of course studies which only give an estimate when no nitrogen fertilizer is administered are not informative for what you would like to investigate.
Good Luck!
-------------------------------------------
Ehsan Motazedi
Wageningen UR
-------------------------------------------
------Original Message------
This message has been cross posted to the following eGroups: Statistical Consulting Section and ASA Connect .
-------------------------------------------
-------------------------------------------
Rose Hart
Recent Graduate
-------------------------------------------
I would like to do a meta-analysis to find an over-all expected yield for switchgrass when no nitrogen is applied and a meta-regression for yield with the co-variate applied nitrogen rates. A few studies give an estimate of expected yield when no nitrogen is applied. Most studies give both an estimate of expected yield when no nitrogen is applied and and estimates when different rates of nitrogen are applied.
The meta-analysis references I've found so far use within-study differences to estimate an effect size and use regression to help explain the estimated between-study variance using a covariate, a covariate not used to estimate effect size.
Please point me to a reference that instructs on how to perform a meta-analysis to find a mean expected yield and then do regression with nitrogen as the covariate. I am assuming this is possible!
Thank you,
Rose Hart