Stepwise Regression

Поділитися
Вставка
  • Опубліковано 3 гру 2024

КОМЕНТАРІ • 106

  • @pedrocolangelo5844
    @pedrocolangelo5844 Рік тому +4

    Professor Obi is definitively one of the greatest on UA-cam. I'm really glad that there are people like you in the world.

    • @VSP4591
      @VSP4591 Рік тому

      You are right. Excelent presentation.

  • @mashaeldewan469
    @mashaeldewan469 7 років тому +35

    Thank you soooo much. I am a PhD student and stat was about to drive me CRAZY. please keep posting, your videos are GREAT.

  • @SATaylor618
    @SATaylor618 6 років тому +3

    Thank you!!!! I was searching all over for the difference between forward and step-wise. You are the first one I could find (after over AN HOUR of searching) with a clear explanation. Thank you soooooo much!

  • @4wanys
    @4wanys 3 роки тому +1

    This old video is the only video that made me understand

  • @walete
    @walete 4 роки тому +1

    very informative and i love your analogies with the fishing and not needing too many spices to make your cooking taste good. thank you!

  • @CynthiaMillikin-o5o
    @CynthiaMillikin-o5o Рік тому

    EXCELLENT explanation and example that clearly demonstrated how to conduct stepwise regression! Thank you!

  • @julianerasmus7540
    @julianerasmus7540 6 років тому +1

    Great video. I hadn't done stepwise in over 10 years and just needed a little refresher and this was GREAT!!

  • @papicoco5069
    @papicoco5069 3 роки тому +2

    I'm doing my undergrad research using the SCOR model, besides using AHP, fuzzy set theory and other PhD level tools to analyse the model the stepwise regression is the easiest for someone like me. Sadly, this was not taught to me on my lectures. I want you to know that your video made me increase my conviction to continue doing my research! All the love man!!

  • @renkano1
    @renkano1 4 роки тому +2

    It was super lucky for me to have met this video before conducting regressions

  • @abderahmanrejeb4423
    @abderahmanrejeb4423 6 років тому +2

    you have a magic style in explaining and making things so clear Myriad thanks

  • @magorzatarams6763
    @magorzatarams6763 5 років тому +3

    Thank You, this is the best explanation of this topic I've seen

  • @Mr_Coolcat
    @Mr_Coolcat 4 роки тому +2

    Brilliant and calm presentation, helped me a lot! Thank you so much!

  • @music_loong
    @music_loong 6 років тому +1

    You are the greatest teacher ever.

    • @PatObi
      @PatObi  5 років тому

      Wow, thank you!

  • @gulaycevik1363
    @gulaycevik1363 5 років тому +1

    Thank you very much, Pat! I am finally able to understand stepwise regression!

  • @tuyizerejeanbaptiste5633
    @tuyizerejeanbaptiste5633 5 років тому +2

    Very easy to understand after hours struggling to. Thank you!

  • @dengdengkenya
    @dengdengkenya 6 років тому +1

    Thank you for sharing this lecture. Great and concise explanation of stepwise regression!

  • @Lucykanya
    @Lucykanya 7 років тому +1

    Many thanks for sharing this - couldn't have been clearer. I should have started here!

  • @lashlarue7924
    @lashlarue7924 2 роки тому

    great explanation - I needed a refresher, thanks!

  • @shayshaylacher
    @shayshaylacher 4 роки тому +2

    Yayyyyyy! This was excellent! PhD student as welllllll! Hi five!

  • @valeriasantos1109
    @valeriasantos1109 4 роки тому +1

    Thank you so much! This explanation was great! You won a new fan! =D

  • @connorjoyce1257
    @connorjoyce1257 6 років тому +1

    Very helpful! Thank you for the explanatory video!

  • @thelateknights
    @thelateknights 3 роки тому +1

    This was so clear. Thanks so much!

  • @tiaemilie8
    @tiaemilie8 5 років тому +1

    Thank you so much for this, found it very helpful for my stats assignment!

  • @jackhenry4785
    @jackhenry4785 5 років тому +2

    thank you .you've made it simple and understood

  • @softwerksaol
    @softwerksaol 3 роки тому +1

    Very nice, just added this to my students' reading list!

    • @PatObi
      @PatObi  3 роки тому

      Thank you.

  • @anomasudrashani3630
    @anomasudrashani3630 4 роки тому +1

    Excellent. learn a lot easy to follow. Thank you.

  • @senioroduro1005
    @senioroduro1005 6 років тому +1

    Bro, You are the best!

  • @NotImmortal
    @NotImmortal 4 роки тому +2

    Brilliantly explained, makes it much easier to implement this on python haha, thank you m8!

  • @jordan2172
    @jordan2172 5 років тому

    Fantastic explanation! Thanks for making it easy to understand :)

  • @shilpagarg9308
    @shilpagarg9308 4 роки тому +1

    Thank you for explaining it well. I have a query if you can be a help. if we are having Rsquare value in 1st step of stepwise regression and having a t-test table for 4 step regression table , how could I calculate R-square value for step 2.

  • @pratikojha4016
    @pratikojha4016 3 роки тому +1

    This is a very good presentation. Thank you for this. Still, I have one confusion, why you didn't see the p-value in every step? Why only in step 4?

  • @ms_1918
    @ms_1918 4 роки тому +1

    Great explanation Thanks

  • @asmaaabdulhamid2292
    @asmaaabdulhamid2292 4 роки тому +1

    it was very helpful, thanks a lot.

  • @AdityaDasika
    @AdityaDasika 4 роки тому

    Agree with the use of t-stat and p-value, but shouldn't we also check the adjusted R-square before adding further variables?

  • @manasatilakraj1289
    @manasatilakraj1289 6 років тому +2

    Why do see the p value in step 4 and not in the previous steps.
    How did you conclude that they are not ststistically significant??

    • @PatObi
      @PatObi  6 років тому +1

      If p-value < alpha, the coefficient is significant. If p-value > alpha, the coefficient is not significant.

  • @wumj2412
    @wumj2412 5 років тому +1

    Very clear explanation thank you

  • @Pickett1312
    @Pickett1312 Рік тому

    Brilliant! Thank you

  • @sandranino5140
    @sandranino5140 5 років тому +1

    AMAZINGGGGGGGGGGG. NOW I CAN DO MY HOMEWORK FOR ANALYTICS

  • @sabinaixchelgarciamendoza
    @sabinaixchelgarciamendoza 4 роки тому +1

    BEATIFUL!!

  • @OnlyTeaGuru
    @OnlyTeaGuru 7 років тому +1

    thanks for the explanations!! Very helpful videos. Regression in SPSS > Regression in Sas

  • @magdalenagutierrez7260
    @magdalenagutierrez7260 8 років тому +3

    Thank you this was a great explanation.

  • @aakupsp
    @aakupsp 4 роки тому

    I have a question. I am trying to find the correlation between multiple signals. Is this a good technique to apply to find which signals are flat signals and present very little correlation between other signals?

  • @billaspiel
    @billaspiel 5 років тому +1

    Hi Obi,
    Thanks for this tutorial . I have one query .
    Do we need to check if p-value < alpha(0.05) in step 1,2,3 as well ?
    if that is the case then From the list of avaible p-values ( p-value < alpha) we need to choose the highest absolute tvalue in it .
    is my understanding right ?
    because only in the step 4 i could see all pvalues > alpha (0.05)

    • @PatObi
      @PatObi  5 років тому +1

      Yes, you go for the highest absolute t. But also, the p-value corresponding to that t has to be less than alpha.

  • @chertify
    @chertify 2 роки тому

    Thanks for a well explained video! I hope you can also share us your csv/excel file so that we can work on it too and experiment.

  • @orkundalyan625
    @orkundalyan625 21 день тому

    Professor Obi, Although the sample count is full in analyses such as t-test and ANOVA, the sample count is 2 less in regression analyses. Is there a reason for this?

  • @dimitris303
    @dimitris303 5 років тому +3

    Dear Obi I found you video extremely helpful! Is it possible to share also your presentation?
    Thnx in advance!

    • @PatObi
      @PatObi  5 років тому

      Thanks. I can send a pdf copy.

    • @dimitris303
      @dimitris303 5 років тому

      @@PatObi you are great! Sincere thanks once again! My email is anastasioud[at]aueb[dot]gr

  • @peanutbrittle7577
    @peanutbrittle7577 6 років тому

    Hello sir @7:26 of the video, why did you select X4? It's t- value was the least actually because it's 'negative' 2. Please correct me if I were wrong, shouldn't that be X1 '

    • @PatObi
      @PatObi  6 років тому +1

      Please refer to 1:56 minute of the video. It's based on the ABSOLUTE VALUE of the t statistics.

    • @peanutbrittle7577
      @peanutbrittle7577 6 років тому +1

      Pat Obi oh yes its absolute thank u so much sir

  • @luiszmaciasvalade5807
    @luiszmaciasvalade5807 7 років тому +1

    Muy claro, gracias

  • @empiricistsacademy7181
    @empiricistsacademy7181 6 років тому

    Awesome video, thanks!

  • @cyborgxxi
    @cyborgxxi 8 років тому +1

    interesting approach!

  • @raymondip9354
    @raymondip9354 4 роки тому

    Hello Dr Pat Obi, thank you for your great efforts and sharing. May I ask for a question? At Step 2, may I know how did you calculate the X5+X1 for b2(0.001), t-statistic(1.422), and p-value(0.162)? I tried to sum up the numbers by X5+X1 from Step 1, but I couldn't get the same numbers. May I know how should I calcuate them correctly? Thank you for your advice in advance. Regards, Raymond from Hong Kong and Macau

    • @PatObi
      @PatObi  4 роки тому +1

      Thanks for the question, Raymond. Each step is a separate regression, independent of the regression in the previous step. In this step, the regression is: Y = B0 + B1X5 + B2X1. The coefficients and stats are obtained from running this regression.

  • @EcoresolveInc
    @EcoresolveInc 7 років тому +1

    Great video!!

  • @funnykid_purab4020
    @funnykid_purab4020 4 роки тому

    great explanation...Thank you...you are DUDE...................

    • @PatObi
      @PatObi  4 роки тому

      You're welcome!

    • @funnykid_purab4020
      @funnykid_purab4020 4 роки тому

      @@PatObi could you please create a video on qq plot ?

  • @flyskypear
    @flyskypear 7 років тому +1

    very clear. Thank you.

  • @sebaragashi8157
    @sebaragashi8157 4 роки тому +1

    Thany you so much !!!!!!!!!!!

  • @reajulchowdhury8534
    @reajulchowdhury8534 6 років тому +1

    Very helpful

  • @bayarsaikhanbaasanbat9009
    @bayarsaikhanbaasanbat9009 5 років тому +1

    I like the analogies: comrade & spice

  • @nileshaqua
    @nileshaqua 7 років тому +1

    Awesome !! Boiler UP :)

  • @r.a.5625
    @r.a.5625 6 років тому

    Hi. So, are we doing the same process for all control variables such as demographic variables we include in the model? Thank you.

    • @PatObi
      @PatObi  5 років тому

      So sorry for late response. In my opinion, the process should apply to all variables. However you should retain any variables you feel particularly confident should be retained in the final model regardless of what the process suggests. This is because a regression model should have sound logic and theory. Refer to the final part of the video.

  • @derrik-bosse
    @derrik-bosse 6 років тому

    What made you decide to go with a four variable model as opposed to a say, 3 variable model?

    • @PatObi
      @PatObi  6 років тому

      Thanks Derrik, for your Q. The final model - shown in the 8th min - was simply based on the forward-selection criterion, described earlier in the 2nd min.

  • @luyennguyen-ri9kp
    @luyennguyen-ri9kp 5 років тому

    How about the logistic regression.Can we use this way to find the best logistic regression model?

    • @PatObi
      @PatObi  5 років тому

      I'm not sure Van. As you know, the logit model stimates probabilities using max likelihood. Perhaps it might be more suited for the linear probability model which estimates probabilities using OLS.

  • @yuhanpan5834
    @yuhanpan5834 6 років тому

    Great video! I have a couple questions. 1) when you said X5+X2, does it simply mean the two numbers adding together? Or do you multiply X5 by the coefficient, and then add X2? 2) How do you determine Beta 0?

    • @PatObi
      @PatObi  6 років тому

      X5+X2 simply means X5 and X2

    • @yuhanpan5834
      @yuhanpan5834 6 років тому

      @@PatObi Thank you so much for your respond. I watched the previous videos and it became much more clear.

    • @fedor6600
      @fedor6600 5 років тому

      @@PatObi hi prof. pat, how do i get the beta coefficient 0 or B0?

  • @sandyyuumr4644
    @sandyyuumr4644 6 років тому

    how about if the variable is latter ? how can i regression on latter variable ?

  • @peanutbrittle7577
    @peanutbrittle7577 6 років тому

    Sir sorry can I ask another question because I was studying this subject in Udemy; instead of taking the independent variable according to the t-statistic, they take the one with the lowest P-value. I read in the following comments in this video that was possible too. However, in scenarios that both of the independent variable having the same P-value or t-statistic, which one is taken to the basket? Thank you so much

    • @PatObi
      @PatObi  6 років тому

      You could select both, in so far as they are both statistically significant.

  • @TheOraware
    @TheOraware 5 років тому

    Thanks for it, but here in stepwise regression there is penalty used in form of AIC , without this penalty term is it not same as simple Forward selection which is purely a greedy approach?

    • @PatObi
      @PatObi  5 років тому +1

      You could say that!

    • @TheOraware
      @TheOraware 5 років тому

      @@PatObi you mean to say both are correct , i can say this video describing as stepwise regression without AIC penalty term? I think something more detail you could give me for my understanding please. There is difference of stepwise regression and simple forward how can i say this video as describing stepwise regression without objective function AIC?

    • @PatObi
      @PatObi  5 років тому

      @@TheOraware Sorry, I'm unfamiliar with the application of AIC in this rather simple variable selection process. Please check other sources help.

  • @g.r.bayazid1546
    @g.r.bayazid1546 6 років тому +1

    Thank you

  • @yassinwessam3500
    @yassinwessam3500 7 років тому +1

    Thanks a lot

  • @fedor6600
    @fedor6600 5 років тому

    @Pat Obi hi prof. pat, how do i get the beta coefficient 0 or B0?

    • @PatObi
      @PatObi  5 років тому

      By running the regression

  • @richardfamson1182
    @richardfamson1182 4 роки тому

    how did you regress 2 or more variables with y? and also can i get your data pls? thanks

    • @PatObi
      @PatObi  4 роки тому

      If using Excel, place the columns of the independent variables side by side. Sorry, the data are not available for sharing.

    • @richardfamson1182
      @richardfamson1182 4 роки тому

      I see.... Thank you so much!

  • @batrapiyushajay8771
    @batrapiyushajay8771 7 років тому

    Can we get any link which directs us to the dataset used in the video ?

    • @PatObi
      @PatObi  7 років тому

      BATRA PIYUSH AJAY I'm sorry, it's not available for general use. Hopefully though the video is helpful.

    • @batrapiyushajay8771
      @batrapiyushajay8771 7 років тому

      Oh, alright. I have a question. In MATLAB, how does the stepwiselm function work ? Does it only perform forward stepwise or backward stepwise or uses both ?

  • @davidalejandrogarciamendez868

    - 2 would not be the lowest?

  • @yousfoss4367
    @yousfoss4367 3 роки тому

    Clair comme l'h2o de roche. thks

  • @karzatben8828
    @karzatben8828 5 років тому

    @pat obi: can u show logistic reg

    • @PatObi
      @PatObi  5 років тому

      I have complete UA-cam videos on logit regressions. Please visit my channel.

  • @thomasleong3510
    @thomasleong3510 7 років тому

    i think is select the lower p - values

    • @PatObi
      @PatObi  7 років тому +1

      thomas leong: Yes, you can use that criterion too.

  • @hollywood632
    @hollywood632 4 роки тому

    throw x4 into the basket to be with his other friends, x1 and x2. lololo

  • @m35926
    @m35926 6 років тому

    You sound like Gus Fring

  • @lugnaanalog467
    @lugnaanalog467 4 роки тому

    AU RAID