<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <p>Dear Tobias,</p>
    <p>Here I comment only to this particular point:<br>
    </p>
    <div class="moz-cite-prefix">Am 18.08.2021 um 14:12 schrieb Tobias
      Saueressig:<br>
    </div>
    <blockquote type="cite"
cite="mid:trinity-8857a43a-29b5-4e61-a42c-7ec067e382b8-1629288724154@3c-app-gmx-bap16">
      <meta http-equiv="content-type" content="text/html; charset=UTF-8">
      <div style="font-family: Verdana;font-size: 12.0px;">[...]
        <div> </div>
        <div>What puzzels me is the difference in the hetereogeneity
          estimates (Tau², I²..etc.). If I perform the analysis with the
          non-transformed means I get different estimates then the
          analysis with log-transformation. </div>
      </div>
    </blockquote>
    This is correct.<br>
    <blockquote type="cite"
cite="mid:trinity-8857a43a-29b5-4e61-a42c-7ec067e382b8-1629288724154@3c-app-gmx-bap16">
      <div style="font-family: Verdana;font-size: 12.0px;">
        <div>If I use the metamean function in meta (for comparison) it
          shows the same heterogeneity estimates wheter I log-transform
          or not.</div>
      </div>
    </blockquote>
    <p>This I don't understand. Heterogeneity measures should in general
      strongly depend on whether the data are log-transformed or not.
      See attached (fictional and quite extreme) example.</p>
    <p>Best,</p>
    <p>Gerta<br>
    </p>
    <blockquote type="cite"
cite="mid:trinity-8857a43a-29b5-4e61-a42c-7ec067e382b8-1629288724154@3c-app-gmx-bap16">
      <div style="font-family: Verdana;font-size: 12.0px;">
        <div> </div>
        <div>Regards,</div>
        <div> </div>
        <div>Tobias </div>
        <div> </div>
        <div> </div>
        <div> 
          <div> 
            <div name="quote" style="margin:10px 5px 5px 10px; padding:
              10px 0 10px 10px; border-left:2px solid #C3D9E5;
              word-wrap: break-word; -webkit-nbsp-mode: space;
              -webkit-line-break: after-white-space;">
              <div style="margin:0 0 10px 0;"><b>Gesendet:</b> Mittwoch,
                18. August 2021 um 13:59 Uhr<br>
                <b>Von:</b> "Viechtbauer, Wolfgang (SP)"
                <a class="moz-txt-link-rfc2396E" href="mailto:wolfgang.viechtbauer@maastrichtuniversity.nl"><wolfgang.viechtbauer@maastrichtuniversity.nl></a><br>
                <b>An:</b> "Tobias Saueressig"
                <a class="moz-txt-link-rfc2396E" href="mailto:t.saueressig@gmx.de"><t.saueressig@gmx.de></a>, "Meta list"
                <a class="moz-txt-link-rfc2396E" href="mailto:r-sig-meta-analysis@r-project.org"><r-sig-meta-analysis@r-project.org></a><br>
                <b>Betreff:</b> RE: [R-meta] Question on combining means
                with robumeta</div>
              <div name="quoted-content">Dear Tobias,<br>
                <br>
                I don't understand what you mean by "Since in some cases
                I had negative values". Negative values where?<br>
                <br>
                This aside, 'var.eff.size' is for the *variances* but
                SEM is the SE of the means. So it should be:<br>
                <br>
                res <- robu(formula = Mean ~ 1, data = dat,<br>
                studynum = Author, var.eff.size =SEM^2,<br>
                rho = .8, small = TRUE, modelweights="CORR")<br>
                res<br>
                <br>
                although the results are basically the same.<br>
                <br>
                Second, LogMean_SEM doesn't seem to be computed
                correctly in your example and let's not call it SEM
                since it's a variance:<br>
                <br>
                dat$LogMean_Var <- with(dat, SD^2/(N*Mean^2))<br>
                <br>
                Now:<br>
                <br>
                res1 <- robu(formula = LogMean ~ 1, data = dat,<br>
                studynum = Author, var.eff.size = LogMean_Var,<br>
                rho = .8, small = TRUE, modelweights="CORR")<br>
                res1<br>
                <br>
                #backtransformed results from log scale<br>
round(exp(c(res1$reg_table$b.r[[1]],res1$reg_table$CI.L[[1]],res1$reg_table$CI.U[[1]])),2)<br>
                <br>
                gives very similar results as 'res' earlier.<br>
                <br>
                Best,<br>
                Wolfgang<br>
                <br>
                >-----Original Message-----<br>
                >From: R-sig-meta-analysis
                [<a class="moz-txt-link-freetext" href="mailto:r-sig-meta-analysis-bounces@r-project.org">mailto:r-sig-meta-analysis-bounces@r-project.org</a>] On<br>
                >Behalf Of Tobias Saueressig<br>
                >Sent: Tuesday, 17 August, 2021 17:48<br>
                >To: Meta list<br>
                >Subject: [R-meta] Question on combining means with
                robumeta<br>
                ><br>
                >Dear All,<br>
                ><br>
                >I have a question concerning the combination/pooling
                of means. Since most of my<br>
                >studies are correlated I deceided to use robust
                variance estimation method to do<br>
                >this. So metamean function in meta was out because
                of this. So I used robumeta.<br>
                >Since in some cases I had negative values I deceided
                to use log transformed means<br>
                >for pooling. My problem is that the estimates look
                reasonable now with robumeta<br>
                >but the heterogeneity parameters(e.g. Tau, I²) are
                all estimated with zero. This<br>
                >is strange. Does anybody know why that is?<br>
                ><br>
                >Regards,<br>
                >Tobias Saueressig<br>
                ><br>
                >Reproducible example:<br>
                ><br>
                >#Example<br>
                > Author <-c("Carboch et al. (2019)","Stare et al.
                (2015)","Mackie (2013)","Hornery<br>
                >et al. (2007)",<br>
                >"Takahashi et al. ( 2006)","Morante et al.
                (2005)","O'Donoghue, P. and Ingram, B.<br>
                >(2001)",<br>
                >"O'Donoghue, P. and Ingram, B. (2001)")<br>
                ><br>
                > Mean<-c(5.93,4.40,6.42,6.70,5.10,6.40,5.70,4.90)<br>
                > SD<-c(0.67,1.73,1.27,2.20,5.19,1.40,1.00,1.40)<br>
                > N<-c(7,5,9,17,16,12,24,35)<br>
                > SEM<-c(0.25,0.77,0.42,0.53,1.30,0.40,0.20,0.24)<br>
> LogMean<-c(1.78,1.48,1.86,1.90,1.63,1.86,1.74,1.59)<br>
> LogMean_SEM<-c(1.50,3.40,2.72,3.57,6.62,2.59,1.16,1.16)<br>
                ><br>
> dat<-data.frame(Author,Mean,SD,N,SEM,LogMean,LogMean_SEM)<br>
                ><br>
                > #Formulas used<br>
                > #SEM<-SD/sqrt(N)<br>
                > #LogMean<-log(Mean)<br>
                > #LogMean_SEM<-SD^2/(N*Mean^2) => taken from
                metamean function in meta<br>
                ><br>
                > #normalscale<br>
                > res <- robu(formula = Mean ~ 1, data = dat,<br>
                >              studynum = Author, var.eff.size =SEM,<br>
                >              rho = .8, small = TRUE,
                modelweights="CORR")<br>
                > res<br>
                ><br>
                > #logscale<br>
                > res1 <- robu(formula = LogMean ~ 1, data = dat,<br>
                >             studynum = Author, var.eff.size =
                 LogMean_SEM,<br>
                >             rho = .8, small = TRUE,
                modelweights="CORR")<br>
                > res1<br>
                ><br>
                > #backtransformed results from log scale<br>
> round(exp(c(res1$reg_table$b.r[[1]],res1$reg_table$CI.L[[1]],res1$reg_table$CI.U[<br>
                >[1]])),2)</div>
            </div>
          </div>
        </div>
      </div>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <pre class="moz-quote-pre" wrap="">_______________________________________________
R-sig-meta-analysis mailing list
<a class="moz-txt-link-abbreviated" href="mailto:R-sig-meta-analysis@r-project.org">R-sig-meta-analysis@r-project.org</a>
<a class="moz-txt-link-freetext" href="https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis">https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis</a>
</pre>
    </blockquote>
  </body>
</html>