We have a 3 data group.
Two of them have the same SD (standard deviation) values.
When we perform the comparison of all these groups the groups with the same SD have a different level of significance from the third group. Is it possible? Or it is a wrong calculation from my side?
As I can assume - the 1st and 2nd groups with the same SD must be equal and show the same level of significance in comparison with the last (3rd) group. Or not? As I understand groups that have the same dispersion (SD) should have the same difference in comparison to another group. Or not?
1st group - SD 0.98
2nd group - SD 0.98
3rd group - SD 0.77.
1st vs 3rd = P0.05.
Does it depend on the digits in the raw? Or does it depends on the SD of the raw?
I'll very much appreciate all effort to explain this fact.
It is absolutely not clear what your model is. What did you compare? Means?
If you were comparing means, please report n and the mean for each group too, please.
Also, why did you tell us about the 1v3 and 2v3 contrasts only? Are these the only contrasts of interest? Were they planned in advance? Thanks for clarifying.
mean 1 =0.000
mean 2 = 0.000
mean 3 = 0.000
400 cases.
I want to ask: if we have the same SD for groups, it is possible to get a different p with comparing to another group, in terms that we compare the means of these groups?
I figure its a formula comparison? You should check if the PE is normally distributed. Usually its not. Constant optimization to ME=0 is questionable if not normally distributed and will therefore not lead to the best result. Optimization for best RMSE or MAE makes more sense. Also, if you want to optimize for ME=0 think about optimizing only eyes with normal AL to 0 and have the long and/or short eyes not skew your constant in one direction or the other. But in that case keep in mind that optimization for ME=0 does not necessarily lead to the best SD and optimizing for SD might lead to an offset in PE.
Dear Jascha,
Yes, you are right.
I got the main idea, thanks a lot, and I will follow your recommendation.
Could you please enlighten us, in what kind of comparison of means you will find a "significant" DIFFERENCE if both mean values are exactly 0.000?!? It doesnt make any sense to me.
Dear Rainer,
Thank you so much for your attention and attempt to help.
It's a hard and cumbersome part, and maybe in the future I'll explain it and return to this theme. But not now.
Anyway, thank you so much.
Rainer Duesing If I am not mistaken, it wasnt a comparison of means, but rather of SD, oriented by this article: Article Review and recommendations for univariate statistical analys...