I measured a variable $X$ and estimated the error bars on my measurement with the objective of testing if $X$ is significantly different from zero. I concluded that $X$ is different from zero only at 2$\sigma$. Following this, I applied the same technique and made use of an (arguably) more precise instrument, and made a second independent measurement. Unfortunately, I concluded that this second measurement is different from zero only at 2$\sigma$ as well.

Is it legit to combine the two consecutive measurements? How can this be done?

  • $\begingroup$ You can easily perform meta-analysis of correlated data with mvmeta in R or Stata. Also consider using individual data meta-analysis though, accounting for clustering. $\endgroup$ – Joe_74 Mar 10 '19 at 13:49
  • $\begingroup$ Since you have the original data why not just do a single analysis with measurement technique as a possible covariate? Or have I mis-understood your design? $\endgroup$ – mdewey Mar 11 '19 at 14:54
  • $\begingroup$ By covariate do you mean control variable? Or do you mean something different? theanalysisfactor.com/confusing-statistical-terms-5-covariate Note that the objective is to get a global picture (and potentially reach a higher level of significance) using the meta analysis. Will including a covariate help? $\endgroup$ – pedrofigueira Mar 12 '19 at 2:21
  • $\begingroup$ I mean a variable which distinguishes the two measurement types. $\endgroup$ – mdewey Mar 12 '19 at 11:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.