Reporting results of Tukey-Kramer posthoc test

#1
My question is probably not the typical query this forum receives, but hopefully there are some researchers on here who can help me.

I actually need to know how you write the results of the Tukey-Kramer posthoc test in a paper.

I completed a one-way ANOVA, followed by Tukey-Kramer at 95% significance. I have my minimum significant difference (MSD) and two of my pairs of variables exceed the MSD and are therefore significant. Now how do I say that in writing?

For example, if I had used student's t (which I know I can't with multiple variables, just bear with me) I would write something like “the results were significant (t = 2.09, df = 23, P = 0.048).”

So what goes in the parentheses for the Tukey-Kramer?

I did a literature search and found a handful of papers that had Tukey-Kramer in it, but they all simply cited it as “(Tukey-Kramer, P<0.05)” and did not report any actual values. Is this an acceptable method for reporting the results of Tukey-Kramer? It seems to me that actual values that quantify the level of significance would be more appropriate (and would be what I would prefer) but I have been unable to find an example.
 

bugman

Super Moderator
#2
I did a literature search and found a handful of papers that had Tukey-Kramer in it, but they all simply cited it as “(Tukey-Kramer, P<0.05)” and did not report any actual values. Is this an acceptable method for reporting the results of Tukey-Kramer? It seems to me that actual values that quantify the level of significance would be more appropriate (and would be what I would prefer) but I have been unable to find an example.
This is the convention I use, and others in my discipline. Usually however, I would report the overall ANOVA results (Fdf,=x; P=y) followed by (for example) X was significantly larger than Y (Tukey HSD; P=0.001) (Table z).

I think you are on the right track! :yup:
 
#3
Thank you for the reply, but I must admit that now you have me a little confused. In your example:
X was significantly larger than Y (Tukey HSD; P=0.001) (Table z).
You indicate a P-value for Tukey. How do you calculate a P-value for Tukey-Kramer and is it even necessary?

I thought P-values weren't necessary for Tukey-Kramer (which is why I assumed all the papers I found didn't cite a specific P-value, only the test limit).

For example, if I have done my Tukey-Kramer at 95% signficance and determined that my MSD was 1.94 and the actual difference for one pair of means was 2.78 than the difference between those means is significant. My original question related to proper conventions for describing that in a paper (e.g. Tukey-Kramer MSD=1.94, Actual=2.78).

Should I be reporting an actual P-value as well to indicate the level of significance?

And if so, how do I calculate a P-value for Tukey-Kramer? (That is a new one for me).
 

bugman

Super Moderator
#4
Didnt mean tro confuse you.

1) by hand, Im not fasmiliar with the formula. Good statistics books should have the method. Or one of the mathematically inclined statisticians on this forum might lend a hand.

2) I am only giving you examples from my disipline (aquatic ecology). We report p-values and then we will present confidence intervals either in a table or graphically.

3) as for conventions for papers. These vary dependiong on the journal AND the disipline. You need to look at specific journals in your areas for this. Again, I obnly gave an example of how I would be doing it if I was presenting these data from my work.

P:D