This video demonstrates how test the normality of residuals in SPSS. The residuals are the values of the dependent variable minus the predicted values.
Thank you for the video! how can I know when to use Shapiro Wilk or Kolmogorov for the residuals? Is it also based on the sample, like if we would conduct a parametric test?
Hi, thank you for the video. If there are points that fall slightly outside -3 to 3 on the x-axis or y-axis when testing for the assumptions of independence and homoscedasticity, what do we do?
Thank you for the video! Around 5:05 you note that no point is outside +/- 3. I was just wondering...what is the advise protocol is value are found that are above/below +/-3? Thanks!
If the residuals are not normally distributed then what can we do to normalize the residuals on Independent Variable? What transformation can we do and how? Thank you!
I saw that your predictor values are all in continuous. Can we apply categorized values too as independent? I have my dependent variable continuous and independent ones are mostly categorized and some of them continuous.
I have done this several times and I still do not have the Shapiro Wilks portion on my table. The only thing that exists is the Kolmogorov-Smirnov. What am I doing wrong?
Just a little side note. A mean of 0 is not the definition of standardized residuals. Non-standardized residuals also have a mean of 0. Standardized residuals are defined by the raw residuals divided by the standard deviation. Nevertheless, good video.
Is it alright for me to go ahead and use multiple regression analysis, if normality of residuals is okay? When I test for this, my normality results turn up insignificant (which is great), but when I test for normality on my raw data it turns up very significant (which is not so great). Any help would seriously be appreciated! I am really struggling with this. Thanks so much in advance. x
Why are you checking normality also on the standardized residuals? I thought by definition the standardized residuals follow a t-student distribution, therefore wouldn't make much sense to check for normality on those, and that is why it's recommended to check normality on the original residuals.
Todd Grande, if the normality test is done and it shows not normal, which statistical test should we do? if normality is violated can we still continue with regression? Aish
+Aishath Shahyma Depending of the characteristics of the data, an ordinal regression may work: ruclips.net/video/ioNr9o8v5o0/видео.html, or a data transformation may be possible: ruclips.net/video/_c3dVTRIK9c/видео.html.
What is important in regression is the normality of the residual, not the normality of variables (if you have a reasonably large sample size, for example, hundreds or even thousands).
I would have taken twice as many hours to complete my assignments without your videos. THANK YOU so much!
Very informative and helpful video. Thanks a ton !!
Thank you for the video, none of my text books described it in a way that i understood.
Thank you very much, it's very clear and informative
Thank you for the video! how can I know when to use Shapiro Wilk or Kolmogorov for the residuals? Is it also based on the sample, like if we would conduct a parametric test?
THANK YOU SO MUCH! Saved my life! :)
Hi, thank you for the video. If there are points that fall slightly outside -3 to 3 on the x-axis or y-axis when testing for the assumptions of independence and homoscedasticity, what do we do?
Thank you for the video! Around 5:05 you note that no point is outside +/- 3. I was just wondering...what is the advise protocol is value are found that are above/below +/-3? Thanks!
It would be really awesome if you had this available as a pdf of steps to take.
Thank you so much! Such a blessing
If the residuals are not normally distributed then what can we do to normalize the residuals on Independent Variable? What transformation can we do and how? Thank you!
Thank you sir for sharing such useful knowledge. 😍👍🏼
You're welcome!
thank you so much its really valuable information thank you
You are welcome -
I saw that your predictor values are all in continuous. Can we apply categorized values too as independent? I have my dependent variable continuous and independent ones are mostly categorized and some of them continuous.
thank you alot ,,the video was very helpful
How do you correct for non normality?
If I use the same data for multiple regressions, do I have to check the residuals for every regression?
whether this method can be used to check the abnormal return in the stock market?
I have done this several times and I still do not have the Shapiro Wilks portion on my table. The only thing that exists is the Kolmogorov-Smirnov. What am I doing wrong?
i love you dr. G
Just a little side note. A mean of 0 is not the definition of standardized residuals. Non-standardized residuals also have a mean of 0. Standardized residuals are defined by the raw residuals divided by the standard deviation. Nevertheless, good video.
HOW CAN WE NORMALIZE A VARIABLE USING BLOM'S FORMULA?
Is it alright for me to go ahead and use multiple regression analysis, if normality of residuals is okay?
When I test for this, my normality results turn up insignificant (which is great), but when I test for normality on my raw data it turns up very significant (which is not so great).
Any help would seriously be appreciated! I am really struggling with this.
Thanks so much in advance. x
I would also like to know the answer on that point...
Me too!
Why are you checking normality also on the standardized residuals? I thought by definition the standardized residuals follow a t-student distribution, therefore wouldn't make much sense to check for normality on those, and that is why it's recommended to check normality on the original residuals.
Todd Grande, if the normality test is done and it shows not normal, which statistical test should we do? if normality is violated can we still continue with regression?
Aish
+Aishath Shahyma Depending of the characteristics of the data, an ordinal regression may work: ruclips.net/video/ioNr9o8v5o0/видео.html, or a data transformation may be possible: ruclips.net/video/_c3dVTRIK9c/видео.html.
If residuals are not normally distributed even after data transformation? All other assumptions were met.
where did you get the information that not normally distributed data is not relevant for big sample sizes?
Typically large sample sizes do not need to be tested for normality based on the Central Limit Theorem: ruclips.net/video/eQde8nDeE50/видео.html
What is important in regression is the normality of the residual, not the normality of variables (if you have a reasonably large sample size, for example, hundreds or even thousands).