If you're talking about meta-analyzing an effect, sure.

Otherwise, probably not without some strong assumptions testing.

Hope this helps

Matt

You may do meta-analysis based on the random-effects model. The heterogeneity index estimated from meta-analysis will tell you whether you could combine all data together.

If you have the original data, merging the data sets and running one regression model is much preferred over meta analysis. If you are sure that all studies were under the complete same circumstances, you can merge the data sets and run a regression right away, e.g.:

lm(biodegrad ~ 1 + temp, data = myData)

If you doubt that the studies are completely comparable (e.g. humidity was not controlled for), then you can still merge, but you should create a categorial variable for the study of origin and introduce that as an effect to the model. It may also be necessary to introduce an interaction effect:

lm(biodegrad ~ 1 + temp + Study + temp:Study , data = myData)

Since you have quite many studies you could also model Studies as a random effect:

library(lme4)

lmer(biodegrad ~ 1 + temp + (1 + temp|Study), data = myData)

This multilevel model gives you estimates for the between-study variance of intercept (1) and temperature slope (temp). When variance is very small, you can drop Study and go with the first linear model.

More about (multilevel) linear models and interpreting interaction effects in my online book:

https://www.researchgate.net/project/Book-New-statistics-for-the-design-researcher

I agree with Martin - with the original data you can use an IPD meta-analysis approach - probably as a multilevel regression with study as a random factor. This is superior in its ability to handle predictors at the individual response level.