Performing a Multiple Regression analysis using JMP including backwards selection model-building steps and constructing a residual plot to confirm assumptions.
Thank you, I couldn't remember how to get the plot for checking the last 2 conditions for regression and with 15 minutes until due date you saved my grade from a hit after I had tried JMP's site and multiple other videos.
Thanks for putting this up, without this video it would have taken me forever to do my stats assignment. For some reason the professors at my university expect us know how to use JMP without ever showing it to us (this is for an intro course).
You can use the stepwise option on the "fit model" screen and it's very helpful when you want to quickly add/remove different variables and see their effect on r-squared.
The math behind the RSquare will cause it to increase as covariates are added to the regression model -- even if the coefficient for that predictor is not significant. So, the RSquare will be maximized when all available covariates are included in the model, even if most of them are only related to the outcome by chance. Note that the "RSquare Adj" adjusts for the number of covariates in the multiple regression model and is a more accurate presentation of the incremental increase of information.
Thanks - useful to see a video on how to do this. I would find it useful to discuss how to avoid over-fitting the model. e.g., if you have 200 rows and 100 different categorical values for each model variable, when do you need a training set and test set and how do you use Jmp to split those out?
@neilarchie1974 You are actually exactly correct - but keep this in mind. Since this is multiple regression, the interpretation of the parameters is a bit more complicated. So ... you can't just say, as weight increase, body fat decreases (negative estimate). You have to consider that we have other parameters in the model, so would have to say for two people with the same abdomen, wrist and forearm measurements, the person with higher weight has lower body fat.
... hopefully this makes some sense since if two people have the same abdomen circumference, but one is heavier, they're probably heavier b/c they have more muscle. Looking at the important covariates this makes some sense that they are areas that don't get bigger if you have more muscle (unlike thigh circumference which could be either muscle or fat).
I hope that you can do more videos like this! JMP should pay you to have this included in their program because their tutorials suck! Thank you so much!
Great video... Thank you. However, I find one aspect of your justification VERY confusing. The RSquare value typically represents a "goodness of fit" (closer to 1, better the fit). Yet in your method of removing non-significant parameters reduces the RSquare value -representing a lesser quality of fit. Is the intent to remove non-significant parameters, or to get a better fit? Thanks again.
Simply phenomenal !! You give a lot of highly accomplished professors and analysts a run for their money. To the point and very well explained.
Thank you, I couldn't remember how to get the plot for checking the last 2 conditions for regression and with 15 minutes until due date you saved my grade from a hit after I had tried JMP's site and multiple other videos.
Thanks for putting this up, without this video it would have taken me forever to do my stats assignment. For some reason the professors at my university expect us know how to use JMP without ever showing it to us (this is for an intro course).
VERY VERY VERY HELPFUL!!! FELT DEFEATED AND STUCK!!! YOURE THE BEST!!! VERY CLEAR
You can use the stepwise option on the "fit model" screen and it's very helpful when you want to quickly add/remove different variables and see their effect on r-squared.
Thanks , Your explaonatoin was very helpful to understand how to work with JMP and also understand the multiple Regression. Thanks a Ton
The math behind the RSquare will cause it to increase as covariates are added to the regression model -- even if the coefficient for that predictor is not significant. So, the RSquare will be maximized when all available covariates are included in the model, even if most of them are only related to the outcome by chance. Note that the "RSquare Adj" adjusts for the number of covariates in the multiple regression model and is a more accurate presentation of the incremental increase of information.
Thanks - useful to see a video on how to do this. I would find it useful to discuss how to avoid over-fitting the model. e.g., if you have 200 rows and 100 different categorical values for each model variable, when do you need a training set and test set and how do you use Jmp to split those out?
Thank you so much for the video for the video.
I was wondering how to use both nominal and continuous variables in a same regression
@neilarchie1974 You are actually exactly correct - but keep this in mind. Since this is multiple regression, the interpretation of the parameters is a bit more complicated. So ... you can't just say, as weight increase, body fat decreases (negative estimate). You have to consider that we have other parameters in the model, so would have to say for two people with the same abdomen, wrist and forearm measurements, the person with higher weight has lower body fat.
... hopefully this makes some sense since if two people have the same abdomen circumference, but one is heavier, they're probably heavier b/c they have more muscle. Looking at the important covariates this makes some sense that they are areas that don't get bigger if you have more muscle (unlike thigh circumference which could be either muscle or fat).
Thank you for the video! It's really helpful. Besides, I love your voice a lot!
I hope that you can do more videos like this! JMP should pay you to have this included in their program because their tutorials suck! Thank you so much!
The video is very helpful! thanks!
Thank you. This video was very helpful. Can you show how to use multiple linear regression with the Stepwise method?
Very useful, thankyou
Great video... Thank you. However, I find one aspect of your justification VERY confusing. The RSquare value typically represents a "goodness of fit" (closer to 1, better the fit). Yet in your method of removing non-significant parameters reduces the RSquare value -representing a lesser quality of fit. Is the intent to remove non-significant parameters, or to get a better fit? Thanks again.
great! thanks.
An advice, pls add the example file to test what did u do on the video or indicate if the file is a sample file from jmp.
Yes.
still helpful in 2018. my professors didnt teach me that!
where can we get the final model in formula to export outside JMP