Announcement

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

  • Random effect logistic regression

    Hi everyone

    I have recently started to use Stata, and confused about how I should interpret the results I have in examining the factors that influence use of private versus public health care. Please assist.


    Random-effects logistic regression Number of obs = 8,033
    Group variable: pid Number of groups = 5,593
    Random effects u_i ~ Gaussian Obs per group:
    min = 1
    avg = 1.4
    max = 4
    Integration method: mvaghermite Integration pts. = 12
    Wald chi2(9) = 775.75
    Log likelihood = -3806.026 Prob > chi2 = 0.0000
    ----------------------------------------------------------------------------------------
    health carew | OR Std. Err. z P>|z| [95% Conf. Interval]
    -----------------------+----------------------------------------------------------------
    Employed |
    Unemployed | .2987269 .0211104 -17.10 0.000 .2600888 .3431049


    communicable |
    Non-communicable | .399234 .0426754 -8.59 0.000 .3237724 .4922835


    African |
    Coloured | .5134034 .0624029 -5.49 0.000 .4045734 .6515086
    Asian/Indian | 1.652477 .4927756 1.68 0.092 .9210933 2.964608
    White | 22.84537 4.908871 14.56 0.000 14.9933 34.80961


    Married |
    Living with Partner | .3180426 .054397 -6.70 0.000 .2274565 .4447052
    Widow/Widower | .5167898 .052537 -6.49 0.000 .423429 .6307355
    Divorced or Separated | .388571 .0889345 -4.13 0.000 .2481138 .608541
    Never married | .3107787 .0315136 -11.53 0.000 .2547637 .3791096
    -----------------------+----------------------------------------------------------------
    /lnsig2u | .8547385 .1373949 .5854494 1.124027
    -----------------------+----------------------------------------------------------------
    sigma_u | 1.533219 .1053282 1.340074 1.754201
    rho | .416755 .0333966 .3531097 .4833012
    ----------------------------------------------------------------------------------------
    LR test of rho=0: chibar2(01) = 158.75 Prob >= chibar2 = 0.000



  • #2
    Besides ctrl v is there a way in which I can paste the result table to be more clear?

    Comment


    • #3
      Yes.
      Code:
      CODE delimiters: see the # toggle in Advanced editor
      Kind regards,
      Carlo
      (Stata 19.0)

      Comment


      • #4
        Click image for larger version

Name:	HC.PNG
Views:	1
Size:	56.6 KB
ID:	1479061


        Which variable should I look at for overall significance? Does the results further indicate that belonging to a race group other than African (except for the Coloured race) increase the odds of using private health care multiplied by 1.6 for Indians and 22.84 times for Whites. Do I read the odds from values under "z"?

        Comment


        • #5
          Pearl, Carlo meant that you should paste the results in code delimiters. We ask that people do not post screenshots, as they are not always readable.

          To answer your question, the global test that all the coefficients are zero is the Wald chi sq statistic and p-value in the top right. The likelihood ratio (LR) test at the bottom tests this model versus an identical model with no random intercept (hence, rejection means that a random intercept is justified; it's like testing that the variance of the random intercept is zero).

          Your interpretation about the odds of using private health care relative to African race is correct. If you're already familiar with regular logistic regression, the coefficients are interpreted the same way.

          The table above doesn't display the odds anywhere. Z is the z statistic. I assume you used the , or option to display odds ratios. If you must have the log odds, you would just omit the option, but log odds are harder to interpret than odds ratios. For that matter, odds ratios are already hard to interpret, so you may want to familiarize yourself with the margins command. (The link goes to some material by Richard Williams, a frequent poster here.)
          Be aware that it can be very hard to answer a question without sample data. You can use the dataex command for this. Type help dataex at the command line.

          When presenting code or results, please use the code delimiters format them. Use the # button on the formatting toolbar, between the " (double quote) and <> buttons.

          Comment

          Working...
          X