0

I'm not one to search for the most tenuous significant difference I can find, but hear me out.

I have some count data with four groups (3 of these can be combined to one, if necessary), groups A, B, C, and X.

Looking at the means and interval plots, X is clearly greater than the others (in terms of mean value), yet I cannot find any statistical test to back this up. This is, I believe, somewhat due to a high variability within groups and the large number of zero values.

I have tried normalized, removing zeroes, parametric, non-parametric, and more, with no success!

Any advice would be greatly appreciated as to how to approach this.

Many thanks.

The link below has the raw data. Groups A, B, and C can be combined into one group if it is relevant.

https://drive.google.com/open?id=0B6iQ6-J6e2TeU25Rd2hsd0Uxd2c

Andrew Cooke
  • 51
  • 7
  • What do you mean by "X is clearly greater"... greater in what criteria? – alpereira7 Oct 23 '17 at 14:37
  • Hi, thanks for the response. The mean of X is 2.82 whilst the means of A, B, and C are 1.75, 0.67, and 1.25. Also, the upper SD intervals for A, B, and C, all fall below the mean of X. P.s. sample sizes for A, B, C, and X are 60, 60, 60, and 140, respectively. – Andrew Cooke Oct 23 '17 at 14:49
  • If X is larger on the whole for your purposes, then forget about any statistical tests. Statistical signficance is a substitute that was invented for problems in which there was no clear assessment of practical significance. If X is larger for practical purposes, then statistical signficance no longer plays any role. – Robert Dodier Oct 23 '17 at 18:33
  • Thanks @RobertDodier, I was having that thought, but it's for a PhD thesis and thought I best do some stats for safety. I'm certainly one to not fuss about with unnecessary stats and perhaps, in this case, no stats is the best choice! – Andrew Cooke Oct 23 '17 at 18:44

0 Answers0