Announcement

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

  • New to stata and need help with matched pair analysis

    Hi,
    I have just started understanding and usung stata, and having used SPSS- I am fumbling a lot.
    I have a gender matched dataset for ADHD vs non-ADHD kids. I matched on gender because the analysis required was simple correlations and associations using chi square/fisher's exact and t-tests and gender was significantly different among the two groups. The ADHD sample is N=23 only while non-adhd N=248 which makes a multivariate analysis a little difficult.
    I want to look at certain physiological and pulmonary outcomes among these two groups. The study is not case control either.
    I have read an article on BMJ, stating that matched analysis is not mandatory after matching and that standard analysis can also be performed. (http://www.bmj.com/content/352/bmj.i969)
    Therefore, I went ahead and performed the standard required analysis.
    However, now I am double minded whether that was the best approach and whether matched analysis is the right way to proceed.
    I am new to stata, I have been meaning to learn it and I believe this can be a start.

    Can you please guide me which commands to use to run t-tests, chi squares/fisher exact, and correlations for matched data?

  • #2
    You'll increase your chances of a helpful response if you follow the FAQ on asking questions - provide Stata code in code delimiters, Stata output, and sample data using dataex.
    For example, you refer to "the standard required analysis" - this may be meaningful to folks in your area, but it tells most of us nothing about what you did. I can think of 20 ways to analyze these data.

    The real issue is an analysis design issue rather than a Stata issue per se. The folks on this list often volunteer help on the design issues, but recognize the distinction.

    If all you wanted was how to run a ttest etc., then the rest of your discussion was irrelevant. If you go to the Stata command line and enter findit ttest, you'll find the link for a t-test. You should go to the subject index (index in the pdf documentation, subject) and look up your tests. Also, if you click statistics at the top of the window and select exact statistics, you'll see a bunch of options there. Once you've done this, then you may have a more specific question we'd be happy to answer.




    Comment


    • #3
      I am new to stata, I have been meaning to learn it and I believe this can be a start.
      Surely, the Stata Forum is a great starting point.

      The Stata Manual is a must read.

      You may access all the content just by asking for help in the command window.

      For example:

      Code:
      . help ttest
      Will lead you to a brief review of the command. If you click on the link in the upper left corner, you will open the PDF file of the Stata Manual.

      The Stata Bookstore , particularly the books on Stata, are also an excellent source.

      The list is endless, you see. But you may also take a look at the Stata Youtube Channel as well as the Stata Blog.


      Hope that helps.
      Best regards,

      Marcos

      Comment


      • #4
        The Stata Manual is a must read.
        Well, I'd add a little specificity to that. In fact, I wonder if anybody has read the entire manual. Maybe Bill Gould, I guess.

        But everybody, no exceptions, should read the Getting Started [GS] and User's Guide [U] portions. They review the fundamentals of working in Stata and cover the commands that are used day-in and day-out in basic data management and analysis. Even they are something of a hefty read, but the time invested will be amply repaid by the familiarity you gain with the most important Stata commands that everybody needs to be able to use. You won't remember all the details, but you will retain enough to figure out, in most situations, which commands are relevant to your problem. And then you can always go to the help files, or the command-specific manual sections for a reminder of the details.

        Comment


        • #5
          In fact, I fully agree with Clyde. My sentence in #3 was not clear enough.

          What I meant by saying the Stata Manual is a "must read" was even less than reading both "start-up" files, but just reading the (wished) excerpt after typing "help" command, then getting access to the topic by clicking on the link of the specific PDF text in the manual.

          I wanted to recommend to always try to do that before searching "outside" the manual because, oftentimes, this first step is just what is needed to accomplish the desired task.
          Best regards,

          Marcos

          Comment


          • #6
            Thank you Phil, Marcos and Clyde for your response. The information provided is very useful and hopefully it will help ease my discomfort related to learning a new software on my own.
            Phil, my design and analysis related question would be that, in your experience, is it okay to use (let's say) independent sample t-test on gender matched data? We have ADHD kids (n=23) and non ADHD kids (n=300). We want to see how their pulmonary function tests differ from each other. Gender was the only variable that was significantly different among the two groups and that is why I decided to match on gender (1:4 matching). Now since n=23 is a small number, I was not comfortable performing a multivariate analysis and wanted to stick with simpler methods instead (Fisher's exact/chi square, t-test).
            after matching, according to the article I cited, we can proceed with independent sample t-test and chi square/fisher's exact. It makes sense for the data we have.
            What will you recommend?

            Comment

            Working...
            X