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!
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!!
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.
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?
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?
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)
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 '
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
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.
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.
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.
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.
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 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?
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?
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 ?
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
![Logistic Regression [Simply explained]](http://i.ytimg.com/vi/C5268D9t9Ak/mqdefault.jpg)
Professor Obi is definitively one of the greatest on UA-cam. I'm really glad that there are people like you in the world.
You are right. Excelent presentation.
Thank you soooo much. I am a PhD student and stat was about to drive me CRAZY. please keep posting, your videos are GREAT.
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!
Great video. I hadn't done stepwise in over 10 years and just needed a little refresher and this was GREAT!!
EXCELLENT explanation and example that clearly demonstrated how to conduct stepwise regression! Thank you!
very informative and i love your analogies with the fishing and not needing too many spices to make your cooking taste good. thank you!
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!!
you have a magic style in explaining and making things so clear Myriad thanks
This old video is the only video that made me understand
Thank You, this is the best explanation of this topic I've seen
It was super lucky for me to have met this video before conducting regressions
Brilliant and calm presentation, helped me a lot! Thank you so much!
You are the greatest teacher ever.
Wow, thank you!
Thank you very much, Pat! I am finally able to understand stepwise regression!
Thank you for sharing this lecture. Great and concise explanation of stepwise regression!
Thank you so much! This explanation was great! You won a new fan! =D
Many thanks for sharing this - couldn't have been clearer. I should have started here!
great explanation - I needed a refresher, thanks!
Yayyyyyy! This was excellent! PhD student as welllllll! Hi five!
Very nice, just added this to my students' reading list!
Thank you.
Excellent. learn a lot easy to follow. Thank you.
Very easy to understand after hours struggling to. Thank you!
Very helpful! Thank you for the explanatory video!
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.
Great explanation Thanks
This was so clear. Thanks so much!
Thank you so much for this, found it very helpful for my stats assignment!
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?
Bro, You are the best!
Agree with the use of t-stat and p-value, but shouldn't we also check the adjusted R-square before adding further variables?
Brilliantly explained, makes it much easier to implement this on python haha, thank you m8!
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?
thank you .you've made it simple and understood
Fantastic explanation! Thanks for making it easy to understand :)
it was very helpful, thanks a lot.
Very clear explanation thank you
thanks for the explanations!! Very helpful videos. Regression in SPSS > Regression in Sas
Dear Obi I found you video extremely helpful! Is it possible to share also your presentation?
Thnx in advance!
Thanks. I can send a pdf copy.
@@PatObi you are great! Sincere thanks once again! My email is anastasioud[at]aueb[dot]gr
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?
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)
Yes, you go for the highest absolute t. But also, the p-value corresponding to that t has to be less than alpha.
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 '
Please refer to 1:56 minute of the video. It's based on the ABSOLUTE VALUE of the t statistics.
Pat Obi oh yes its absolute thank u so much sir
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
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.
How about the logistic regression.Can we use this way to find the best logistic regression model?
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.
great explanation...Thank you...you are DUDE...................
You're welcome!
@@PatObi could you please create a video on qq plot ?
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.
What made you decide to go with a four variable model as opposed to a say, 3 variable model?
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.
Brilliant! Thank you
AMAZINGGGGGGGGGGG. NOW I CAN DO MY HOMEWORK FOR ANALYTICS
Hi. So, are we doing the same process for all control variables such as demographic variables we include in the model? Thank you.
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.
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?
You could say that!
@@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?
@@TheOraware Sorry, I'm unfamiliar with the application of AIC in this rather simple variable selection process. Please check other sources help.
how about if the variable is latter ? how can i regression on latter variable ?
how did you regress 2 or more variables with y? and also can i get your data pls? thanks
If using Excel, place the columns of the independent variables side by side. Sorry, the data are not available for sharing.
I see.... Thank you so much!
Thank you this was a great explanation.
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?
X5+X2 simply means X5 and X2
@@PatObi Thank you so much for your respond. I watched the previous videos and it became much more clear.
@@PatObi hi prof. pat, how do i get the beta coefficient 0 or B0?
@Pat Obi hi prof. pat, how do i get the beta coefficient 0 or B0?
By running the regression
Can we get any link which directs us to the dataset used in the video ?
BATRA PIYUSH AJAY I'm sorry, it's not available for general use. Hopefully though the video is helpful.
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 ?
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
You could select both, in so far as they are both statistically significant.
interesting approach!
Great video!!
BEATIFUL!!
Awesome video, thanks!
- 2 would not be the lowest?
I like the analogies: comrade & spice
Awesome !! Boiler UP :)
very clear. Thank you.
Muy claro, gracias
Thany you so much !!!!!!!!!!!
Very helpful
@pat obi: can u show logistic reg
I have complete UA-cam videos on logit regressions. Please visit my channel.
Thank you
i think is select the lower p - values
thomas leong: Yes, you can use that criterion too.
Thanks a lot
Clair comme l'h2o de roche. thks
throw x4 into the basket to be with his other friends, x1 and x2. lololo
You sound like Gus Fring
AU RAID