A Gentle Introduction to ANOVA - The Problem of Probability Pyramiding (12-1)
HTML-код
- Опубликовано: 10 апр 2017
- Why do we need ANOVA? Can’t we just compare multiple groups with lots of little t tests? Unfortunately, when we use multiple t test, the error rate skyrockets. The effect of probability pyramiding kicks in and we begin making Type I errors with no way of telling which findings are wrong. The ANOVA allows us to compare means from 3 or more groups by first analyzing the variance between and within groups, then doing a follow up (post hoc) test.
This video teaches the following concepts and techniques:
Introduction to the ANOVA concept
Link to a Google Drive folder with all of the files that I use in the videos including ANOVA datasets. As I add new files, they will appear here, as well.
drive.google.com/drive/folder...
Table of Contents:
00:08 - Statistics for the Flipped Classroom
02:13 - Mo’ Measures, Mo’ Problems
02:32 - Why Not Use Lots of t Tests?
03:10 - Lottery Analogy
04:29 - Probability Pyramiding
05:25 - Solution? …ANOVA !
06:56 - One Way ANOVA 
Хобби
That's so true:
"Lottery is a tax on people who are bad at math and don't understand probability."
Well said!
Fair enough, here's another way to put it: "Lottery is one of the ways of exploiting those who have found themselves in a precarious situation, or those who know that if they lose their job or mess up, they will find themselves in a precarious situation."
If you genuinely think people do lottery just because, or even predominantly because they don't understand math, you know nothing about psychology, and you should refrain from posting comments on youtube. Btw, you posted it only to positively distinguish yourself from "those others, who are not as wise as me", just so you're aware. Good day.
Can you PLEASE do a video for 2 way ANOVA, Repeated Measures ANOVA, MANOVA, ANCOVA and a break down of the ANOVA table, using variance to compare means (within vs between/f ratio etc) and related measures etc. PLEASE! You are literally the only person on RUclips that I've been able to learn stats from....and I've spent weeks on here looking. Test this Thursday, and then the final in 3 weeks, for grad NP stats. begging!
Sorry for not having those videos yet. I have most of those coming up next semester but, unfortunately, not soon enough to help you with your final. If you want to reach out to me through my Missouri State University email, I will send you some notes. Hope that will help. Good luck on your final!
😂😂😂 “Mo’ Measures, Mo’ PROBLEMS”❤️❤️❤️💛💚✅✅😂😂😂😂 #truth #truthakins 🙏🏽
Thank you so much for this expanded explanation on ANOVA
The p=0.05, when I reject the null hypothesis five times, why the alpha=0.4? should it equal 1-（0.05^5)??
You literally saved my Statistics exam last year. Passed it with a 9 (1=bad 10 is super good)
Excellent! So glad to hear it.
your explanations are awesome, but maybe I need more information for my project... do you know of any good sources for other kinds of ANOVA/GLM?
thank you for the deeper understanding of the subject, I think I could not do anything without it!
The best book I can recommend for learning the varieties of GLM is Andy Field "Discovering Statistics with SPSS". Excellent explanations. Even a used copy of an earlier edition (3 or 4) will have everything you need.
@@ResearchByDesign thank you 🙏🏻
Do you have anything on two-way ANOVA? Your videos have been sooooo helpful!
I do not have anything on two-way (or factorial) ANOVA right now, but I do have something in the video production pipeline. Probably will not get there until summer, however. In the meantime, I recommend Laerd Statistics for some excellent tutorials. They can definitely help. Good luck
@@ResearchByDesign Thank you for sharing your videos!! I will definitely look that up.
Hello , I do not get it. How did you calculate that runing t test in 5 groups has .40 alpha error. each 2 group has 0.05 . so if we add all 0.05, it would be0.5 not 0.4.
I can't replicate the formula here because of the limitations of the comment box, but the math looks like this: m = 10; 1 - (1 - .05)^10 = .40. The m value (which becomes the exponent) is derived from the number of groups. Best I can do here in the comments...hope it helps.
Hey sir I’m not sure if this will make any sense but I REALLY hope you can help as it’s my last chance of getting this right. I am researching the mean weight of twelve difference brands of vitamins. Two of those brands are pharmaceutical grade and ten are food grade. From the data I obtained I can see that the variability within and among the food grade vitamins is huge and as expected little variation between the pharmaceutical grade. I would like to display the inter and intra-variability of the 10 against the 2 using statistical tests but I’m so confused as to which statistical test I would use on this. PLEASE PLEASE help me sir :(
You can use an F-test for inferences between two population variances. The simplest way is divide the variance of the larger sample variance by the smaller sample variance (that is your F value) and the use an F table (such as for ANOVA) with n-1 degrees of freedom for the numerator (larger) and n-1 degrees of freedom for the denominator (smaller). You would need the variability for all pharmaceutical and all food grade, each as their own group. That is a lot to explain in the comments sections, but I hope it helps
What analysis should I use to compare GPA and preferred learning style of students (VAK) ?
Assuming that GPA is a scale variable and VAK is a nominal (categorical) variable, you would use an independent t test to compare 2 VAK groups and a one-way ANOVA to compare 3 or more VAK groups.
logistic regression
3:18 teacher made me laugh 😂 now I love studying Anova🤗😂😂
I have to compare impact of Alcohol consumption on few variables of a memory test (5 variables)? Can i use independent T test 5 times or ANOVA? Thank you
That one is a little tricky. You COULD use t-tests, but multiple t-tests lead to Type I errors, so the alternative would be to use ANOVA. However, if you want to simultaneously test 5 dependent variables, you would use a MANOVA, which is a much more complicated test. On the other hand, if your five variables are all scale measures of memory, you could combine/average them and use that as a single DV in a t-test. Good luck
i am doing knowledge and attitude regarding pain management survey to a group of student. dependant variable is pain management and my independant variables are knowledge and attitude. pls advice on analysis method...
Sounds like all of your variables are scale level. If you want to predict pain management from knowledge and/or attitude, that is a regression. You would need categorical IV's to do an ANOVA/MANOVA.
RStatsInstitute thank you
Hi Prof, could you make a video on Two-Way (Factorial) analysis with medical data please?
That is a great idea...I lectured on two-way (i.e. factorial ANOVA) this week but have yet to develop a clever example to illustrate the techniques. The medical data would be a good starting place. Thanks!
@@ResearchByDesign Thanks Prof for your prompt response. Eagerly waiting for your upcoming video. I have been totally dependant on your video lectures for my statistical analysis.
Which test to analyse relationships between tourism and GDP
I am guessing correlation
Absolutely correct...you would use correlation, assuming that both variables are scale level (i.e continuous; interval or ratio)
Hello. In my study I want to find the association between personality trait (continuous variable, independent variable ) AND adherence to medication ( categorical , dependent variable) please advice on which analysis to use
)
That is the setup for a logistic regression. You use your continuous IV to predict the belonging in the dichotomous DV categories (adhere vs. not). Look for "Laird Statistics" for a wonderful step-by-step for how to do it in SPSS
That is a point biserial correlation. Same math as a t-test, but structured to measure the association between a scale variable and a dichotomous variable (compliant or not). I don't have a video yet, but Laird Statistics has great tutorials that explain how to do it in SPSS.
ON AVERAGE, one will be a type I error. It isn't a certainty for any given analysis
Good point...on average you will get a Type I error somewhere and you can't be certain which of the tests is that error. Thanks for the comment!
For the variable Age I have classified as 5 categories :
Below 25, 26-35, 36-45, 46-55, Above 55
But SPSS is showing only 45-55 and Above 55 categories..
Any kind of analysis I do I am facing the same problem..
What could be the solution to this ?
Hmmm...I would need to see your data to diagnose the problem. I could be that you are missing a code for the lower categories, perhaps you have a select cases filter turned on, they could be entered as string variables instead of numeric and need to be recoded. Could be any of a number of things. Sorry I could not be more helpful.