Announcement

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

  • Skillings-Mack test

    I recently had the opportunity to use a Friedman type test, and was very happy to find an implementation of the Skillings-Mack test in Stata by Mark Chatfield and Adrian Mander in the Stata Journal. Many thanks to the authors for their effort, as I had both ties and missing data, which their skilmack program handles.

    The implementation of the Friedman test cited in Chatfield's paper doesn't seem to handle ties as well as the implementation of the Friedman test available in R. The R version gives a statistic closer to the skillmack test than does the Friedman implementation in Stata, when ties are present. Apparently, formulas for the Friedman test are often reprinted that assume no ties are present.

    https://jimgrange.wordpress.com/2014...ith-tied-data/

    <https://en.wikipedia.org/wiki/Friedman_test> note the comment about ties following the formulas give here

    When I apply the friedman test in Stata and R on the data from the Grange post, as well as the skilmack test, I get:

    Stata Friedman test statistic and p-value of 13.2821 and 0.2753

    R friedman.test statistic and p-value of 7.6 and 0.02237

    Stata SM test statistic and p-value by simulation: 7.717 and 0.0190

    ​​​​​​​-Dave
Working...
X