Been way too long since I have done analysis of variance, Bayesian theorum or meta-analysis. Sorry, but I would need a refresher. I do miss my epidemiology days [sigh]...
Hi guys,
I'm currently taking a Biostatistics course and the assignment I'm working on deals with ANOVA and the Bonferroni Correction. Does anybody have a really good grasp on the Bonferroni Correction method and would care to dumb it down for me. If I understand correctly it's a method to determine if differences between groups is due to random chance or if there is actually a significant difference. In this particular example I'm working on it's three 'treatment groups', all with the same sample size of 44, and I've already gone through the ANOVA method and determined that I'm strongly rejecting the null that all three groups are equal.
Already tried to find a Khan Academy video on it but no such luck.
Thanks for anybody who can help.
Goalies: If I'm pickin em you best be sittin em!
Been way too long since I have done analysis of variance, Bayesian theorum or meta-analysis. Sorry, but I would need a refresher. I do miss my epidemiology days [sigh]...
@SmittysRant
Basically you are running a Bonferroni to be sure that you are not finding significance due to the number of variables. If you want to hit a p value of 0.05 then you need to divide that by the number of possible paired comparisons (so in your case 3 - AvB, AvC, BvC), giving you a target of 0.017 to be reached between any one comparison.
It's been too long since I've had to actually do this at a serious level though so someone else should probably check that I'm right (good chance I made a mistake). Doesn't Dobber have a stats degree or something?
Ah thanks for chiming in doulos. I've gotten as far as dividing my p-value by 3 and I know I"m supposed to carry out a t-test for each individual comparison but I have no idea what it means lol. I really wish there was a Khan Academy vid for this stuff...that guy's videos are incredible and really focus on helping you understand why you're doing certain calculations and what they tell you which makes it easier to problem solve down the road.
Thanks Bomm as well. If anyone else wants to chime in that would be great.
Goalies: If I'm pickin em you best be sittin em!
Side story: The engineering curriculum I took at U Waterloo had a statistics class in 2nd or 3rd year. I tried to get through the entire class without opening the text book on my overrated (/incorrect) thought that my math was good enough to stand alone.
Ha, FAIL. I think I came out of that class with a grade in the 60s. (As compared to the 90s I pulled off in high school).
Anyways, we used a text book called "Miller & Freund's Probability & Statistics for Engineers". Fifth Edition. 1965/1977/1985/1990/1994
571pgs.
I keep it in my office at work.
Pg.415 it has about 1pg on Bonferonni.
If you like, I can scan it through our printers and email it to you.
Sometimes, just reading an explanation from another source can help you understand it better.
(But... it's now been about 15 years from that class for me... and, like I said, I didn't really ever crack the book... so I'd be lying to try to offer any help. Good for you for taking a stats class though!!!!)
I think you'll find it's easier than you think once it "clicks".
You have 3 groups and you've figured out the P value between each paired comparison right?
So:
AvB = 0.016
AvC = 0.9
BvC = 0.025
Without adjusting with the Bonferroni you would see significance in comparisons AvB and BvC if you were looking for a value of 0.05. However, since we are using 3 total paired comparisons we need to adjust. Another way to look at it is to multiply the P value by 3 (3 total paired comparisons. So...
AvB = 0.016 * 3 = 0.048
AvC = 0.9 * 3 = 2.7
BvC = 0.025 * 3 = 0.075
After adjusting with the Bonferroni method you've now lost signficance in the third group (BvC), but still have significance (0.048 < 0.05) in the AvB group.
Not sure if that makes it clearer or not.