@@dataexplained7305 May I ask some Qs.. with your example where you have 14 variables to choose from, how would you select or how would you know which ones to start with. Also how will we know the model selection is "homoscedasticity enough" so we can stop looking further. I'm no stats background here and not sure how exactly to look at the 4 plot charts to see it's "pretty enough" lol. (my current QQ is rather a straight line but the first plot Alteryx put a curvy line through like an ugly earthworm lol)
oh I should add.. so if my current model is not "pretty enough" (not homoscedasticity), is my only option keep removing and adding something back to re-do my regression? So potentially I may need do to C(14,2) or even C(14,2), C(14,3), C(14,4)..attempts until I finally find the best model with the prettiest plot?
@@MonsieurSchue you might want to first check using pairs() to see which of the features are correlated and then move on to regress them.. like you said, you can introduce new features in and useless features out to make the model better
Looking for a guidance tool to create large data testing tool UAT vs PROD..is that something you have a video on? Basically big data excel sheets comparison
Data validation comparing the source data numbers to see if they match the target numbers ? UAT being source and target being PROD.. You can get this done using Alteryx or even excel itself(if data volume is smaller) Give it a try, if not, I can do one for you !
@@livelovelough3 I know, that's the pain with excel. Definitely this can be done with any other powerful engines, even Tableau can help. I will do a video on this one soon.. Keep in touch..
Thanks, love the video, simple and straight forward, clearly explained! (but the audio is kinda horrible..)
Definitely some scope for improvement.. lol
@@dataexplained7305 May I ask some Qs.. with your example where you have 14 variables to choose from, how would you select or how would you know which ones to start with. Also how will we know the model selection is "homoscedasticity enough" so we can stop looking further. I'm no stats background here and not sure how exactly to look at the 4 plot charts to see it's "pretty enough" lol. (my current QQ is rather a straight line but the first plot Alteryx put a curvy line through like an ugly earthworm lol)
oh I should add.. so if my current model is not "pretty enough" (not homoscedasticity), is my only option keep removing and adding something back to re-do my regression? So potentially I may need do to C(14,2) or even C(14,2), C(14,3), C(14,4)..attempts until I finally find the best model with the prettiest plot?
@@MonsieurSchue you might want to first check using pairs() to see which of the features are correlated and then move on to regress them.. like you said, you can introduce new features in and useless features out to make the model better
Re-uploaded with upgraded audio !!
That’s awesome explanation. Do you do anything on logistic reg?
I have a video on Logreg using R.. statistically it's same in alteryx too..
@@dataexplained7305 thanks do you have anything to share on stock market price predictions etc.,?
@@dataexplained7305 can i have the link?
@@girimpr ruclips.net/video/mnpYGWrAJw4/видео.html
Looking for a guidance tool to create large data testing tool UAT vs PROD..is that something you have a video on? Basically big data excel sheets comparison
Data validation comparing the source data numbers to see if they match the target numbers ? UAT being source and target being PROD.. You can get this done using Alteryx or even excel itself(if data volume is smaller)
Give it a try, if not, I can do one for you !
DataExplained yes correct.. excel takes forever as the data volume is high .. and data is coming from Uat and tableau environments for diff reports
@@livelovelough3 I know, that's the pain with excel. Definitely this can be done with any other powerful engines, even Tableau can help. I will do a video on this one soon..
Keep in touch..