How do I meaningfully display Kruskal Wallis Data when I have a lot of zeroes?

I'm comparing ordinal variables (first is location, second is source) to the number of mites per bee. The problem is that several of the samples don't have mites; others have a TON of mites. My data isn't normal (right-skewed). So I decided to use a Kruskal Wallis test followed by a Dunn's test for pairwise analysis. Now, I do get significance, but my graph looks incredibly wonky and I don't think it is particularly useful to the viewer. I think since the majority of samples had no mites, the median is zero for all my groups. BUT there is a difference between groups. So my questions are: Do I just scrap this graph and use the table? Would changing the numbers to ranges help? that way the ranking process is easier? Please show a better display if you think there is one.

Lineage & mite load ?2(3) = 5.02, p = .170, indicating that mite loads were similar for each level of Lineage Source & mite load ?2(2) = 21.22, p < .001, indicating mite loads were significantly different between levels of Source Region & mite load ?2(2) = 69.51, p < .001, indicating mite loads were significantly different between levels of Region

  Score Mean Difference SE Z p-value Adj. p-value
S-HI 76.66 9.33 8.21 <0.001 <0.001
S-UT 55.75 14.03 3.97 <0.001 <0.001
UT-HI 20.89 14.17 1.47 0.14 0.42
  • This is not my area of expertise but you may want to allow more time.


Answers can be viewed only if
  1. The questioner was satisfied and accepted the answer, or
  2. The answer was disputed, but the judge evaluated it as 100% correct.
View the answer
  • @Matchmaticians. This is the best I can answer the questions and concerns raised.

  • I will answer all concerns and raised questions in my answer.

  • @ Rage: Could you please address Sizzledee concern below?

  • Thank you Rage.

  • @ Sizzledee: Rage has revised the solution. Please review the revised answer and see if you find it satisfactory.

  • Sounds good, thanks.

The answer is accepted.