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

  • Choice of the statistical test

    Dear statalist members,
    Please I need your advice on how to choose the best test for the following scenario
    I have the baseline echocardiographic measures, the patients had intervention and the echo at the last follow up was reported.
    I need to study the effect of the procedure on the changes on echo measures but I've only pre-operative echo, echo immediately postoperative and at the last follow up.
    Can I deal with these data as panel data? Can I use mixed effect to analyze this with only 2 measures for each patients at different time points (last follow up ranges from 3 months to 3 years)?
    Thank you

  • #2
    welcome to this forum.
    As per your description, you can use either -xtreg- or -mixed- (assuming that the regressand is continuous).
    With no more details I find difficult to reply more positively.
    Kind regards,
    (Stata 16.0 SE)


    • #3
      Hi Carlo,.
      Thanks for your reply.
      Is it valid to do longitudinal analysis with only 1 or 2 measures for each participants?
      Please find the screenshot of the data attached.
      We performed a procedure on a heart valve and we have the pressure pre-procedure. We need to study if there is an effect for the procedure on the pressure during follow-up
      we have the pressure recorded after 2 weeks of the procedure then the last measure. The timing of the last measure is different in each patients.
      Would you please elaborate more on how I can reshape these data before xtreg?
      Is it valid to use mixed linear effect?

      Best regards,
      Click image for larger version

Name:	screenshot.jpg
Views:	1
Size:	217.2 KB
ID:	1536784


      • #4
        you can run -panel data analysis if (and only if) you have two or more measurements (also known as waves of data or data waves) of the same variables on the same sample of patients.
        Otherwise, you can only go -regress- (which is not in line with your research goal).
        Go to -xtreg- entry in Stata .pdf manual and/or take a Comprehensive look at any decent textbook on panel data econometrics.
        As a relevant aside, please do not post screenshots (as they are often unreadable and Always impossibkle to elaborate on) but posta what you typed and what Stata gave you back via CODE delimiters (see the FAQ). Thanks.
        Kind regards,
        (Stata 16.0 SE)


        • #5
          Hi Carlo,
          Thanks again for your reply.
          There is a question for me regarding panel data which I couldn't find a satisfactory reply to it. I appreciate if you give me suggestion:
          In medical studies especially the retrospective ones, repeated measures are usually collected at different times.
          For example in patient 1, echocardiography was performed at baseline then after 1 y, 2.5 y, 4 y, 5.2 y, then for patient 2 the measures were taken at baseline then 0.5 y, 1.7 y, 3 y, 3.5 y, 4.8 y, and so on
          When reshaping data and setting j(years), it deals with y1, y2, …. as if they are measured at the same time.
          how can we consider different timing when dealing with data?
          I'll really appreciate your response


          • #6
            this issue is actually quite frequent also in econoic evaluation performed alongside clinical study (especially observational ones). Mostly depends on the allowed dispersion around the scheduled follow-up time (eg, at 6 month +/- 2 months) and/or on patient severity at the enrollment (the more severe the patient at baseline, the more frequent the follow-up).
            The usual fix is to consider visits instead of years and then go panel at 1st, 2nd...n visit: obviously, this fix is sometimes unsatisfactory.
            Kind regards,
            (Stata 16.0 SE)