2
$\begingroup$

If not, since a systematic review would be the next option, are there any analyses you can conduct on the results of the studies that one intends to analyze

$\endgroup$
  • $\begingroup$ Can you edit this to clarify exactly what you mean since as @bjorn points out what you describe is exactly what meta-analysis does. $\endgroup$ – mdewey Dec 31 '18 at 12:01
2
$\begingroup$

I do not know what precisely you mean by raw data, but let us assume $a=1,...,N$ articles each conducting a study using there own dataset $D_a$. They each try to estimate the same theoretical parameter $\theta$ - for example the wage returns to education - then you could simply build a taxonomy for the variables and regress estimates $\hat \theta_a$ on dummies for which type of variables they had included in their models when estimating $\theta$. This would only require information on which type of variables they include in their statistical model and the estimates they arrived at but would not include the original data used in the articles $D_a$. Here is a concrete example of such a study from the field of transportation economics concerning estimation of agglomeration effects metastudy example

$\endgroup$
  • $\begingroup$ Thank you for the advice. But what if the studies are looking for trends within data. For example, if each study took a sample of people with diabetes, and looked at different aspects of their lives (diet, region where the live, etc.) and from that data, each study found some factors that they believed were correlated to causing diabetes, could you run an analysis solely on the trends that each study found, without taking each studies' raw data? and if so, how? $\endgroup$ – ConstantFuture Dec 30 '18 at 19:13
  • 1
    $\begingroup$ If you find the answer an answer to your question pls. accept and upvote. If not and you find the answer useful pls. just upvote. If you think the answer is insufficient you can edit your question to include the more specific details you give in the comment above or you can open another question. This is not meant in an unfriendly way, just to keep in line with the question-answer form of SE for the benefit of other users as well. The point is simply to keep the essentials in the questions and the answers. $\endgroup$ – Stop Closing Questions Fast Dec 30 '18 at 19:24
2
$\begingroup$

The meta-analysis literature is primarily about doing exactly that. I.e. methods too combine evident be from multiple studies using either using an estimate $\hat{\theta}$ of the quantity of interest and its standard error (that's the standard scenario), or using some other kind of aggregate data that might be reported in journal articles or registries.

$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.