Statistics 101: Logistic Regression Probability, Odds, and Odds Ratio

In video two we review/introduce the concepts of basic probability, odds, and the odds ratio and then apply them to a quick logistic regression example. Understanding these concepts is crucial to knowing what logistic regression does and the ability to interpret computer output after running logistic regression. Thanks for watching! - Brandon
My playlist table of contents, Video Companion Guide PDF documents, and file downloads can be found on my website: www.bcfoltz.com
#statistics #regression #machinelearning

[@hamzamohd.zubair1709]

I have a masters degree, and I don't say this often. Your videos are the best teaching material for Probability and Statistics. You are up there with Sal Khan, may be better.

[@sierrawhiskey5155]

Your videos are an excellent service to humanity, thankyou

[@vylonygo7410]

Besides your undoutedly strong teaching skills, it is your playlist management skills and consistency that hold you at the top for me! Thank you for giving this to us for practically nothing!

[@BrandonFoltz]

I appreciate that! Thank you so much for the kind comment. I am so happy you find the videos helpful. 🤗

[@Teodora4537x]

Great video! Makes me wonder why I'm paying tuition fees when you're doing a better job than my teacher.

[@sureshm430]

THANKS is not enough.. i should say more than that, I started my day with logistic regression, planned to complete in 2 to 3 hours.. i had gone through lot of stuff which lead to confusion between logit and sigmoid function..... felt bad and sad, that i couldn't make it, then your stuff came, it is SAVIOUR.. You explained very clearly, detailed every thing in simpler manner. I ended my day with SMILE with your stuff.. APPRECIATE.. APPRECIATE.. APPRECIATE... Expecting stuff on Machine Learning Algorithms too.

[@MrsTurunTurun]

I just started to watch you videos for fun. I have always thought of taking statistics.. thanks for showing me some basic points and how awesome and interesting it is! Can't wait to enrol in a full time statistics studies!

[@Marta43505]

Thank you so much for these videos. You may be the one who saved me from repeating an awful class :)

[@BrandonFoltz]

*NEW VIDEO* is up! Second in the Logistic Regression Series

[@KehleboeGongloe]

Hi Brandon, I would like to thank you for these selfless efforts in helping millions of students around the global. That said, I would like you to know that there many places on earth today where students cannot even watch youtube for steady 2 minutes without interruption because of poor communication infrastructure. This brings me to question, do have all of your lectures of CDs or DVDs for sale? Or, have you published a book with these introductory lectures? I would like to buy these as I prepare to return to Liberia where I know it will be difficult to watch your lectures on youtube. Thank you again kehleboe

[@MsSKim1]

thank you so much for this i love you

[@kamryncherry9661]

Thank u brandon

[@sedaaydin3000]

I have seen many stats videos so far and i can comfortably say that your teaching technique is the best! thank you so much brandon!

[@ChryslerPTCruiser]

These videos are really helping me with my fundamental understanding of how this stuff works thank you! The warning was super helpful because that part has been confusing me

[@313neo]

Thank you very much. The simple, plain, and coherent explanations are very helpful in understanding a new concept. Your time & effort is very much appreciated.

[@amairaguptasworld4728]

you are simply amazing. How easily you explained Logistic Regression. Great going Brandon.

[@rnd322]

Brandon, your videos are awesome! Such a brilliant way of simplifying all these concepts.

[@david4102]

Brandon, this is really helping me with my thesis. Thank you

[@MrSofuskroghlarsen]

You have a talent, sir. I feel like i understand this very well. As someone who's attention span is very bad, you're very excellent at simplifying the stuff and also at keeping a calm and collected approach. I have learned so much from your statistics 101 videoes! Bless :)

[@amitverma20076]

Thanks for explaining so difficult concept in such a simple way. Great explanation!

[@BasuthkarKiran]

I have never seen logistic regression being taught in such a logical and simple way. Thank you!

[@BrandonFoltz]

Glad it was helpful! You are very welcome. I am grateful for your time.

[@rachelburns6917]

thank you Brandon! I have my research & analytics final in an hour and a half and this is a lifesaver

[@ssslogan9481]

Thank you for your videos. They are comprehensive and easy to catch.

[@fetabrown]

I've been tearing my hear air out with risk ratio vs odds ratio really mean - your meteor/lighting example really helped that make sense!

[@jaimepuccioni9598]

Excellent video! A great intro and/or review of logistic regression.

[@soniaturrini5039]

you are a blessing and will save my academic career. thank you

[@panagiotisgoulas8539]

@ Brandon Foltz 8:53 I think I got it. You compare 2 different body weights: w1 P(sleep apnea)=p1 and P(not sleep apnea)=1-p1 and w2 P(sleep apnea)=p2 and P(not sleep apnea)=1-p2 . So if w1=50kg and w2=51kg (1 unit increase in independent variable) the odds ratio will tell us how a 1kg increase in body weight increase the odds of having sleep apnea. Also P(struck by lighting)=p1 and P(hit by a meteor)=p2, both very low p1>p2 odds(struck by lighting)=p1/(1-p1)>>>odds(hit by a meteor)=p2/(1-p2) General takeaway is what you said 10:10

[@markevans5648]

Thank you so much! You are an extremely talented teacher.

[@alexdamado]

thanks for posting, really cleared up some concepts for me.

[@suneel8480]

Felt in love with your teaching style :)

[@pirasa5]

Great video, thank you! Could someone explain and tell where to find how we calculated the increase in odds to 1.98 with 10 lbs gain and increase in odds to 3.867 with 20 lbs. Couldn't get my head around how these were calculated on the slide at 9:40. Thank you.

[@OubouchouN]

Excellent Brandon. I really love your videos. You saved my session :)

[@caterinaversari9871]

You are going to save me from a bad grade in my exam. Thank you, incredibly helpful!!

[@growingisgood]

Thank you so much. You're explanations are awesome !

[@scaras323]

YES ! This is what I was looking for ! Thank you so much !

[@sherryyu7259]

Thank you very much. Hope Brandon if you can do a lecture in multinomial logistic regression.

[@johnessien2167]

Hi Brandon, thanks so much for your videos. I have been binging on them for a statistics assignment and you have done a wonderful job of breaking stats down. One concern I have with logistic regression would be helpful would be how to select the best logistic regression model. I ask this considering that the assumptions for linear regression are different from the assumptions for logistic regression. Or are the principles and the factors to look out for the same as with linear regression? I would infinitely appreciate your take on this. Thanks for all you do

[@jaci1011]

Thank you soooo much Foltz. I have one comment for those who are confused about the difference between probabilities and odds, and one question. My comment is ' If we partition all the possible outcomes of an experiment into two part, and we call the first part "The outcomes of interest" and the second part "The outcomes not of interest", that is, all possible outcomes=outcomes of interest + outcomes not of interest Then depending on the problem you considering if you interested in the ratio "outcomes of interest/ all possible outcomes" then you talking about probabilities, and if you interested in the ratio "outcomes of interest/outcomes not of interest" then you are talking about odds, "visualisation of these ratios will help a lot in differentiating between probabilities and odds". My question is. 'According to the definition of odds "odds=p/1-p" it is clear that odds are ill defined when "p=1"' could you please elaborate more on this point?'

[@rajabtanbal1]

You are a gifted teacher.

[@texaspolygraph]

Awesome man. You explain things very well.

[@sirginirgin4808]

Excellent explanation. Kindly keep uploading more lectures . Thanks

[@AfiaWerekoaah]

Thanks soo soo much for this vid. I'ts soo simple and soo clear to understand. Much appreciated. With love from Amsterdam, The Netherlands.

[@annamirjam8164]

These are awesome and very helpful!

[@lexmarkx83]

Hi Brandon, You have been very helpful. Thank you so much! I have a request for you. Can you dig deeper into logistic regression and make videos on multiple logistic regression? It seems like almost all of your logistic regression videos are at an introductory level so I thought if you can upload more challenging stuff! Appreciate the effort as always!

[@texaspolygraph]

Thanks again for another excellent video.

[@rachelli6817]

I like everything you say! so so useful!

[@tannaz5724]

How great your explanation 👌🏼thank you

[@afiabakwasi544]

You made my evening, God bless you

[@TheHectormanuto]

Note that log refers to the natural logarithm, as it is the common convention in computer science. The logit function takes as input values in the range 0 to 1 and transforms them to values over the entire real-number range, which we can use to express a linear relationship between feature values and the log-odds

[@geethasubramanian9964]

Good video.....helped a lot in understanding the concepts...Thank you

[@Arendt-Foucault]

Best statistics course ever . ThX

[@jamesawuye5102]

Good work done

[@mohammadmaniat1040]

how great your explanation. thank you thank you

[@gaon5880]

Excellent video. Thank you

[@amanyasharon5227]

You're an excellent teacher

[@mangaart3366]

Amazing video I learned so much!

[@JasleenKaur-xl2fx]

Hi Brandon, Thanks for putting together the core basics and sharing them with all. (y) BIG THUMBS UP I found two small errors in the slides. 1) The warning slide, last bullet .. should mention "CHANGE" - odds can have a large magnitude "CHANGE"! 2) the example you give of being hit by meteor or lightning, cannot have high odds while having lower probability ,but the CHANGE IN ODDS of being hit by one of the two based on say weather conditions or astronomical positions may be of LARGE MAGNITUDE for one with much lower probability.

[@sigridskjnne1132]

Really good and informative video

[@danielkursitis9031]

This was so helpful, Thank you! 😁

[@badoiuecristian]

Great video, thank you !

[@nilanjanmitra6239]

Hi Brandon First I want to thank you very much for bringing every topic alive. I have gone through most of your videos-pretty good. I would like you to present some other topics like non-parametric test, design effects. Why you have not said any word on tests of proportions? its interval estimate and sample size? More you will deliver more we will enjoy Regards - Dr.Nilanjan

[@ALBANIAN4FREDOM]

Brandon, thank you soo much

[@GOURAVKUMAR-tg5gl]

excellent explanation....great

[@zckfu]

Is it possible for you to discuss about the "COUNT DATA" using poisson regression, quasi-poisson, or even the negative binomial regression?

[@elizabethlambert8606]

A legend among men! You may have just saved by Research Statistics grade!

[@meghnamane3375]

Very nice explanation

[@JasleenKaur-xl2fx]

Say the probability of being hit by metoer, on a particular day changes from regular recorded of .005 to .009 which is an increase of 80% while that of being hit by lightning, changes from regular recorded probability of .0005 to .0079 which is an increase of 1480%. To an ordinary person the probability of being hit by meteor(.009) is higher than probability of being hit by lightning(.0079) But the magnitude of CHANGE in their odds presents the real deviation from regular behaviour! :)

[@ozguctakmaz]

Excellent!

[@abhijha3849]

Excellent video...

[@faruqsandi6184]

I love your video. Great.

[@irockrock44]

Thanks for a great video series. However I like to point that in last slide 'WARNING' - I think you meant separation between probability and odds ratio. not odds. when you say odds of getting hit by lightning is higher then odds of getting hit by meteor we are talking about odds ratio not individual odds. Correct me if I am wrong btw as I am a new learner of probabilities.

[@vinayvernekar8331]

Thank you , Thank you for the video..............

[@eduugr]

Thank you !!

[@rohitekka2674]

I binge watch your playlist instead of a Netflix series.

[@purplesubmarine83]

Thanks for this video, really helpful! I just have a stupid doubt: if there is an increase in odds of 1.07 when weight increases of 1 pound, this doesn't mean that the odds is equal to 1.07 when you weight 1 pound more, but just that 1.07 is the difference in odds between the final and the initial weight, right? I have this doubt because you write both 'increase by ...' and 'increase to ...' and that confuses me a little...

[@himanshu431]

Thanks a lot.

[@vinayvernekar8331]

You explain it better than my professor

[@cbd89]

YOU ARE A SAVIOUR !

[@iamdhruvcool]

quality video bro

[@marloon5]

fyi the link to the third video does not work for me.

[@blankpages3881]

Hi. You stated that the odds of an event happening are the same for the same level of increase/decrease i.e., if the odds of having Sleep Apnea are 4 x more with a 10 pound increase from 100 pounds to 110, then this is sacrosanct for any level (200 to 210 or 500 to 510). However, I am unable to wrap my head around the intuition behind this. Wish you covered the intuition or math in this video. Hope the following videos cover this, moving onto the next!

[@somcana]

How do you convert odds of 19.5 to a percentage. I am more familiar with 1 digit and the decimal values

[@emmah8712]

So basically probability is the likelihood of an event occuring while odds simply propense the chances of its occurence. So the chances that some one may experience an event is a function of the changes in odds that is subjected to the variablility of a significant event and the initial probility of him/her experiencing at the time before the change in odds. I hope i am not confused by the way i suscribed. Thanks alot. Please if you dont ind can you do a tuto on spss on binary multiple logistic regression with both nominal and ordinal data set thanks loads.

[@Highlander0689]

This is a good video, but what about the calculation and interpretation of marginal effects at the means for this type of model? Also, what about the similarities and differences between probit and logit models?

[@attitude928]

Any way to get Odds Ratios for the excel linear regression program?

[@rgsoni]

Where is your video #3? I can't find it.

[@chaowang1071]

So much videos, what you expained show me the truth.

[@Streetkillerful]

At the last part of the video, was that odd , an odd ratio you referred to?

[@muhammadelemam346]

thanks for your explanation but there is no soft source you can send it so we can start our training on your great knowledge

[@sajjadabdulmalik4265]

Hi Brandon, what made you choose numerator as loaded coin is there any rule for this?

[@hiCooljacko]

Good video

[@dr.satyabratasahoo5644]

Ravishing

[@bobababy2436]

By any chance could you share the slides as well? Thank you so much!

[@corneliadiana9919]

please make some videos regarding survival analysis

[@wisesoar]

Hi Brandon, Thanks for your video, I really enjoy watching and learning from it. However, I have some questions and confusion about the odds ratio. How does 10 pound increase in body weight increase the "odds" of having apnea to only 1.98? According to the statement, "...the odds ratio for a variable represent how the odds change with a 1 unit increase in that variable while holding all other variables constant", Shouldn't it be 1.07 * 10, thus gives 10.7 increase in the odds? How come it becomes the exponential relationship, 1.07^10 = 1.98? Thanks

[@BrandonFoltz]

wisesoar Hello! The relationship is exponential by definition. Watch the rest of the videos in the series and it should all make sense. :) If you are still confused after watching the rest, let me know and we can go from there. Thanks for watching!

[@wisesoar]

***** Thanks so much for your reply. I figured it out. It's actually related to the logged odds ratio.

[@thej1091]

how is odds ratio explained at 10:43 similar or even remotely close to odds ratio explained in the previous slide?

[@nguyendinhtuan1]

To predict in the parity game on the internet, what algorithm should be applied, sir? I'm from Vietnam.

[@vitus008]

can I get this slides?

[@danguerriero3094]

Will be starting in a month or two your videos query could this help in stock trading. I realize there are no sure bets

[@BrandonFoltz]

+Dan Guerriero Hello! It could be used to convince oneself that there is a hard science behind stock trading. But there is much random walk in the market. I am a proponent of a diversified buy and hold strategy of index funds and ETFs in a tax deferred account with dividend reinvestment. No realized capital gains or transaction costs which usually eat up results in a hurry. But this could be used to develop some interesting models or just explore the market.

[@danguerriero3094]

+Brandon Foltz Brandon thank you for your time. Your right no way to predict which makes diversify sensible. I appreciate your response and scholarship

[@cristtos]

With the weight variable odds ratio - I get a different value for the 10 pound increase - unless I am missing something. 1.07 to the power of 10 is 1.96715135728956532249, which rounds up to 1.97, not 1.98. Am I wrong?

[@LonglongGuitar]

the link in the video 'Click here for video #3' is not working

[@indirakasnawati6029]

hey how did you get the number for loaded coin flip? :(

[@kristinrichie3718]

Hi Brandon, I have a question about the independent variable. So far, you have been using examples were the independent variable is quantitative or interval-ratio. However, can logistic regression be used with an independent variable that is nominal/categorical and binary. I have been told that as long as both independent and dependent variables are binary, logistic regression can be used. Thanks!

[@korneliaekeli7117]

+Kristin Richie I'm not Brandon, but...Yes, I believe it can! You just have to dummy-code the categorical variable, as you would in linear regression when dealing with categorical variables :)

[@kristinrichie3718]

Thanks Kornelia! :)

[@kristinrichie3718]

Also, when you say dummy-code the categorical variable, what exactly do you mean? Are you referring to transforming/recoding the variable into a binary category?

[@korneliaekeli7117]

Yup :)

[@Allyballybean]

Should it say "Increases the odds by a factor of 1.98" ( not increases the odds to ...)