Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Meta analysis of means for one group - recurrent unanswered question

    I want to meta-analyze the mean_bp (continuous variable) across several studies with a single group of patients. Metan and other commands require two groups of patients. Metaprop allows for one group of patients but only deals with binary variables. I am surprised to see this question pop up recurrently but still no answer...

    Thank you,
    Jonathan

  • #2
    Jonathan,

    I may not understand your question, but this would seem to do what you want.
    Code:
    use http://fmwww.bc.edu/repec/bocode/m/metan_example_data
    metan tmean tsd
    Red Owl

    Comment


    • #3
      Thanks Red, the only problem with this is that it does not allow me to specify the study size (tsample) which would seem to be required to properly perform the meta-analysis. I tried: "metan tmean tsd" but it does not work. Thoughts??

      Jon

      Comment


      • #4
        Jon,

        The weighted average in a meta-analysis is based on the inverse standard errors rather than directly on sample sizes, so showing sample sizes is just a matter of displaying additional descriptive information about the studies in a table or forest plot.

        You can add the sample sizes information to the forest plot with the following command plus the -lcols()- option.
        Code:
        use http://fmwww.bc.edu/repec/bocode/m/metan_example_data
        metan tmean tsd, lcols(id tsample)
        Does that do it?

        Red Owl

        Comment


        • #5
          Yes! (I figured it out a few minutes ago, and your response has solidified it)

          The problem that I was having was that I was inputting the standard deviations instead of the standard errors. So I created a variable for the standard errors, re-ran the metan command, and then obtained the expected meta-analysis results.

          Code:
          gen se = sd/sqrt(n)
          metan mean se

          Thanks again,
          Jon

          Comment


          • #6
            Thank you for this topic thus far. I am doing a silimar type of analysis and want to estimate baseline quality of life (scale 0-100) across several cohorts.
            I have several studies reporting N's, means and SD's of patient groups.
            Jonathan Afilalo, you choose mean and SE for in the metan command, why not mean and SD?
            I know the difference between SD en SE, but can not decide which one to use in the analysis.
            Can both be used, but then the interpretation is different? (variation around the mean versus deviation population mean)

            thank you,
            JJ

            Comment


            • #7
              Guys, -metan- undestands that when you have two variables, such as metan var1 var2, the second one is the standard error. Hence, you should specify the standard error, and not the standard deviation.

              Comment

              Working...
              X