The Bonferroni Correction - Clearly Explained

[@skippersmg]

Why did i join 2 years of statistic lectures? Yeah right, to sleep and catch up most of it in couple of minutes, years later

[@StevenBradburn]

Haha! More content to come :)

[@sa-pt3kf]

I like your explanation, its simple and to the point. However, the switching letters make it kind of uncomfortable to watch what is written. Seeing as this is a one year old video and I haven't watched other videos yet I don't know if this is a common effect you use, however I'd advise against it. Otherwise, as I said great video and thanks for the information!

[@sandorattilapabar6773]

I am a simple man. I saw there is a new video on the Top Tip Bio and just clicked on it.

[@Kikkerv11]

Thank you, this is much clearer than my professor explained it!

[@chunt5584]

Thank you for this wonderful video. You helped a completely clueless psychiatry resident about to sound very smart next week :D

[@benbadura3050]

Simple, understandable, to the point! great video Thx!

[@kalday656]

GREAT video! Clear, concise, and made for regular, non-stat individual! Thanks! Now, if you could do that for everything stat (ANOVA/ANCOVA/MANOVA etc)

[@hannaholding7742]

Yes please!

[@phai44]

best explanation on yt. thanks, you are an amazing person

[@senla249]

This is an incredible explanation, thank you so much!!

[@roGCyborg]

This video is excellent! Thank you for condensing the material and assisting me immensely!

[@StevenBradburn]

Glad it was helpful!

[@Steffystr8mobbin]

@@StevenBradburn it was helpful but the moving text effect is really distracting and overexaggerated imho, that aside thanks for the info

[@FearlessFresh]

This was super helpful. Thank you!

[@StevenBradburn]

Glad it was helpful!

[@AB-fm7qh]

Thank you! This was incredibly helpful!

[@Pezonman98]

Great video! Cant wait for the next videos on the other methods like the Tukey correction

[@StevenBradburn]

Thanks Martin. Working on these soon :)

[@김중섭-l5x]

Thanks a lot! so easy to calculate what I've done for 4groups

[@pierre-alexandrequittet6461]

Very clear and very helpful thank you so much !

[@LauraWyattMusic]

Did you do a video for Holm-Bonferroni in the end? I can't find one on your channel. I think it would be really helpful, you have a way of explaining things that makes so much more sense than the other stats channels on KZbin!

[@ma-xt2zc]

Great video! Could you explain the differences between LSD Correction and Bonferroni Correction. Thanks!

[@christinechan821]

Beautiful explanation. Thank you.

[@mikekarypidis]

Thanks for the well explained Bonferroni Correction. You have mentioned that there is a Tuckey's correction. I was looking through your videos and I did not find it. Would you be so kind to post the link. (I use Tuckey' s HSD often as a post-Hock Test in my ANOVA analysis and I would like to know more about it)

[@LauraWyattMusic]

Thank you so much for such a clear explanation 🙏

[@debbiejones2742]

Great, clear explanation. My only comment was that I found the jumpy text hard on my eyes. But otherwise great, thanks

[@StevenBradburn]

Thanks for the feedback Debbie

[@konstantinaavdala5336]

Hi, I loved your videos. Can you please give me some references for bonferroni correction. Thank you in regard.

[@dicao4458]

Great video！but it would be greater if the subtitles didn't jump up and down

[@ThEKeeYs]

that was actually very helpful .. thanks homie

[@collegestatistics]

Better to connect Bonferroni's bound to the binomial distribution to get more precision in your experiment-wise alpha.

[@kaylabrand7403]

Great explanation! However - can anyone explain why we care about family-wise error? If you do 20 hypothesis tests with alpha=0.05, then yeah, there is a good chance that at least one of those tests will falsely reject the null hypothesis. But isn't this what we agree to when we say we are setting alpha at 0.05? Why should type I error be so much more important than type II error?

[@hanyferdinando112]

Thanks for this nice and easy-to-follow tutorial. I have questions... I compared whether Group 1 is smaller than Group 2. To get the data for this comparison, I used a formula with two parameters to play around. Let's say, I have 12 sets of data for comparison, i.e., G1 vs. G2 using scenario #1, G1 vs. G2 using scenario #2, ..., and G1 vs. G2 using scenario #12. I don't want to compare which parameter combination is the best. The main goal is to test if G1 < G2. Should I use Bonferroni correction?

[@elisaschmied911]

Can I also do a Bonferroni correction when I do not have multiple comparisons/ t-test? In my research I just calculated correlations between variables.

[@ravivadali7836]

Awesome explanation. Great work.

[@shortandsweet2767]

The Bonferroni correction is an adjustment made to P values when several dependent or independent statistical tests are being performed simultaneously on a single data set. Please explain what is the meaning of "single data set" there....

[@anasantos9264]

Parabéns pela didática, muito bem explicado. Ganhou uma nova seguidora.

[@amirafredriksson9952]

I loved this, thank you!

[@piersantesestini1607]

Nice video, But what is the difference between m and k?

[@JoshH282]

Great explanation!

[@StevenBradburn]

Glad it was helpful Josh!

[@aaronm9491]

Amazing explanation

[@ranielirag2899]

what if I have 4 groups, then one is the control group(1), and the remaining three(2,3,4) are the experimental group, is it okay to just compare 1 to 2, 1 to 3, and 1 to 4? Considering that I am already done doing anova with the 4 groups.

[@njm_01]

when we are doing the Bonferroni correction should we use that alpha level only for our hypothesis testing or should we also use it as a significance level while we are comparing the sociodemographics of our subjects?

[@sunsite15]

just a quick question. why do we perform only 6 tests at the group comparison. shouldnt it be like3 for each group. group1 vs 2,34. group 2 vs 3,4,1 etc

[@shosho43666]

Can you upload the Code for the graph to see the differences between the Bonferroni correction and without the correction ?

[@minjang4288]

what if one-way ANOVA test give you the p value greater than 0.05 (in my case I got 0.08 with my sample group)? I understand there is no significant difference in means between my groups, but what is the next? how can I procedure Bonferroni correction although no significant difference in means between my groups? what is the alpha value then? can you just do same? Please let me know.... I desperately need your help

[@sohilausamarabie9631]

Thank you for the super clear explanation!! can't thank you enough :)

[@StevenBradburn]

Very welcome Sohila

[@cinthyadelangelzuvirie5066]

Hi. What about the Sidak's method video? I am interested. =)

[@maryamsediqi3625]

Thank you, was very useful 🙏

[@meshachjones6909]

Thank you so much

[@felipegutierre7037]

Thank you!!!

[@niashapompey4371]

My professor said the alpha is divided by K( number of groups) to find the new Bonferroni alpha. Now im confused

[@AndrywMarques]

Great video!

[@christianrodier3381]

Thanks that was helpful!

[@its_RD_]

crystal clear

[@steven9492]

the letteres moving around really madfe it hard to read :(

[@TheHeadincharge]

The more I learn and understand about FWER, the more problematic it becomes in application as it assumes that all of the comparisons resulted in a p-value of .05 which of course will never realistically occur. Using that assumption to make a correction is dubious at best and is especially problematic when comparisons are less than 50. I also think including the “and more” bit can be misleading and potentially wrong since it is only representing the chance that one event occurs. No assumption can actually be made beyond that.

[@mateussantosesilva3899]

Thank u so much!

[@likexu2615]

Brilliant!

[@ParanParan-gf9zy]

How do you get those p values

[@amritsudershan8620]

Thanks ❤🌹🙏

[@jackieng127]

You saved me

[@HakobAvetisyan]

Thanks for the great explanation. I have a question regarding the cross-comparisons of the groups in the example. Even though there are totally 6 cross comparisons, each group is compared with other groups only 3 times (G1xG2. G1xG3, G1xG4. Similarly for G2, G3, and G4). Shouldn't this mean that the m value is 3, instead of 6 ? Thanks for your answer.

[@StevenBradburn]

Thanks for the feedback Hakob. So, there is a lot of debate about what actually makes a 'family' of tests. Generally, people would correct for every individual hypothesis tested. So, you could say G1 vs G2 is 1 hypothesis, G1 vs G3 is 1 hypothesis, and so on and so forth. In total, for my example anyway, this would mean 6 individual hypotheses. I hope that makes sense. Steven

[@iFastee]

@@StevenBradburn are you schizo?

[@khushinotariya6164]

how did you get individual p value for each group?

[@StevenBradburn]

Hi Khushi, You will get an adjusted p-value for each group comparison (test) when performing post-hoc analyses. Maybe this video will help: kzbin.info/www/bejne/Y4u6Y6iJg9KnitE

[@mirianarizzo8928]

Hi! Just a quick question: when you say 'number of tests', what do you actually mean? Indeed, if I have a 10 thousand genes, and I'm doing t-tests for each of them, how many tests should I perform?

[@joefish6546]

You should consider the Benjamini-Hochberg method. It ranks by p-value and then adjusts as it goes, so it will not reduce alpha as aggressively as the bonferroni for the gene pairs with the biggest differences. However, it will still be very small for 10,000 genes. Maybe consider a way to study less genes? Good luck and go cure something!

[@lindapetit-frere9657]

I still dont understand. I thought using this method would create a more possible chance to decrease error 1

[@aradopsis2164]

Great content! If I may I ask, is it correct if I did the correction the other way around? What I did is I multiplied all the p-values with number of observations instead of dividing the alpha with the no. of observations. The interpretation is still the same but I'm not sure if this is "statistically correct"

[@StevenBradburn]

Thanks for the feedback Ara. Yes, of course! In fact, I know SPSS does this to report the adjusted P value. Just be sure to make it clear when you report your results that the P value has been adjusted (via the Bonferroni correction)

[@daurenbekkairat4303]

Thank you a lot! Quite informative video! Also, is it worth discussing the improved Bonferroni Correction? For example improved Bonferroni test introduced by R. J. Simes. Alternatively, will similar modern Bonferroni types outperform current improvement? Reference for an improved Bonferroni: Simes, R. J. 1986. “An Improved Bonferroni Procedure for Multiple Tests of Significance.” Biometrika 73 (3): 751-54. Link: www2.math.uu.se/~thulin/mm/HW2-Simes.pdf

[@StevenBradburn]

Thanks very much for sharing Daurenbek! :)

[@daurenbekkairat4303]

@@StevenBradburn Always welcome! I did not quite get it, however, it's not a problem for you :)

[@laura2364]

Oh god i must be stupid. I really don't understand.

[@Ehsas-rf2dp]

How you get the p value for the 6 groups ?