1
$\begingroup$

I am trying to understand the results on the kruskal wallis test that are produced in R. The data that I am testing:

Class       Branch  LA_type Method_type     Method_call     Branch_type     Branch_condition    Tested_parameter
Goal        12      Smooth  public static   never called    IFNE            TRUE                 String
TreeApp     20      Rugged  constructor     none            IF_ICMPGE  FALSE                     int
Password    4       Smooth  private         never called    IFEQ    FALSE                        int
XMLParser   9       Rugged  constructor     none            IFNONNULL   TRUE                     String
MapClass    33      Smooth  public          never called    IFGT    FALSE                        double

We want to know where there is a difference between, for example, the method call of a smooth LA and the method call of a rugged LA. Note that this is only a sample to be shown here. When I apply the Kruskal-Wallis test with LA_type and Branch_type, I get the following result:

Kruskal-Wallis chi-squared = 33.657, df = 15, p-value = 0.003803

While the result of LA_type and Method_call is:

Kruskal-Wallis chi-squared = 85.377, df = 4, p-value < 2.2e-16

My question is what does the chi-squared mean? How does it indicate the significant difference between the groups? and is it really correct to say there is a significant difference between, for example, the method call of a smooth LA and the method call of a rugged LA?

Improve this question
$\endgroup$
3
  • 1
    $\begingroup$ I find the question hard to follow ... 1. What's the response variable being compared? 2. If you're only comparing two groups why use Kruskal-Wallis? 3. What is it you're originally trying to find out? $\endgroup$
    – Glen_b
    Oct 10, 2019 at 12:39
  • 1
    $\begingroup$ The dependent variable for Kruskal-Wallis should be numeric, or at least ordinal. I'm wondering if you are (incorrectly) feeding the function a nominal dependent variable, and the function is coercing it to a numeric variable? $\endgroup$
    – Sal Mangiafico
    Oct 10, 2019 at 16:04
  • $\begingroup$ About the chi-sq part of your question. In some circumstances, the test statistic $H$ is approximately chi-squared: "If possible (no ties, sample not too big) one should compare $H$ to the critical value obtained from the exact distribution of $H.$ Otherwise, the distribution of $H$ can be approximated by a chi-squared distribution with $g-1$ degrees of freedom." Quoted from a longer explanation in Wikipedia on K-W test. $\endgroup$
    – BruceET
    Oct 11, 2019 at 8:15

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.