Hi there,
I am a medical researcher and am sending this e-mail to ask for help on interpretation of the statistical analyses in my recent study.
I studied the effect of two new drug candidates (denoted as drug X and Y) on the effect of managing a bone disease in an animal model, together with two control groups. So the grouping is as follows:
Group A: control (without any treatment).
Group B: also control, treated with a drug currently used in clinics.
Group C: treated with drug X.
Group D: treated with drug Y.
I used ANOVA followed by LSD test to make comparisons between all the four groups.
Overall, the results showed that: group B is slightly better than A without significant difference (as a pilot study, my sample numbers are quite limited, so not surprising). And C is moderately better than B, without significant difference either; but the difference between C and A is significant. D is very marginally worse than A, moderately worse than B and significantly worse than C.
As I am writing a manuscript for publication, considering the drug Y appears not immediately promising, I am wondering if I need to include D (i.e. only report group A B and C). Then I did another ANOVA with only A, B and C. The program output a p=0.056, so it did not recognize the difference among A/B/C to significant, then the post hoc analyses should be ignored. However, when I actually look into the post hoc results, the difference between A and C has a P of 0.020.
So I am quite confused what the ANOVA told me? If I include the group D, seems I can claim C to have significant therapeutic effect (over group A); but if I drop group D, the significance disappeared (as ANOVA did not OK me to further pursue pos hoc). So, although D did not physically do anything between A and C, it could change out conclusion!
Is this a normal situation in medical statistics? And usually how do any of you deal with this kind of dilemma? Is there any guidelines?
I will deeply appreciate it if any one could offer me a bit hint. This will be a part of my Doctoral dissertation. Please send your comments to [log in to unmask]
Thanks again.
I Livingstone
PS: I tried t-tests across all groups, which all textbooks claimed as not good, and it showed the difference between A and C is significant.
---------------------------------
Sucker-punch spam with award-winning protection.
Try the free Yahoo! Mail Beta.
|