Very informative video! I got a data (sample size more than 100) where the z value of skewness and kurtosis fits into the range of +3 and -3, which suggest the data is normal. But the kolmogorov smirnov test suggests the test is non normal. What inference or conclusion can be drawn from this? is the data normal or non normal?
So what about ordinal data? Scale responses, agree-disagree? How do I make an "appropriate" graph for those? Do I just mark those as scale-type data? You confused me xD
From my understanding, statisticians disagree here. Some say it is OK to use a histogram for ordinal data (see, for example, sphweb.bumc.bu.edu/otlt/mph-modules/bs/bs704_summarizingdata/bs704_summarizingdata4.html#:~:text=When%20one%20is%20dealing%20with,categories%20have%20an%20inherent%20order). Others say you should not use a histogram for ordinal data and should use a bar chart or box-and-whisker plot instead (www.uni-bamberg.de/fileadmin/eng-ling/fs/Chapter_19/411Frequencydata.html).
@@virginiawickline2609 is it normal to remove it. I am about to consider this variable as dependent one. It was behaving like poisson but with this behavior it doesn't respect poisson assumption either.
@@phd.eriolamariuscharlotade5634 Hi! You are reaching the limits of my knowledge. However, it is fairly common to remove cases with high Cook's D values so you are removing the influence of outliers. See, for example: towardsdatascience.com/identifying-outliers-in-linear-regression-cooks-distance-9e212e9136a
Thank you so much for the informative video!!! Truly appreciated it 😍😍
My pleasure - glad it helped!
Very informative video! I got a data (sample size more than 100) where the z value of skewness and kurtosis fits into the range of +3 and -3, which suggest the data is normal. But the kolmogorov smirnov test suggests the test is non normal. What inference or conclusion can be drawn from this? is the data normal or non normal?
Very helpful!
Thank you! very helpful
My pleasure!
Thanku very much mam.. I have a question that how we interpreted it in our article. Can you suggest me any Article in which transformed data is shown
I apologize for missing your comment. I do not have an article on hand to suggest. Sorry!
@@virginiawickline2609 ok.. Thanku very much mam for your reply
So what about ordinal data? Scale responses, agree-disagree? How do I make an "appropriate" graph for those? Do I just mark those as scale-type data? You confused me xD
From my understanding, statisticians disagree here. Some say it is OK to use a histogram for ordinal data (see, for example, sphweb.bumc.bu.edu/otlt/mph-modules/bs/bs704_summarizingdata/bs704_summarizingdata4.html#:~:text=When%20one%20is%20dealing%20with,categories%20have%20an%20inherent%20order). Others say you should not use a histogram for ordinal data and should use a bar chart or box-and-whisker plot instead (www.uni-bamberg.de/fileadmin/eng-ling/fs/Chapter_19/411Frequencydata.html).
@@virginiawickline2609 You rock! Thanks a lot. Appreciate the sources too :)
super helpful
Glad to hear - thanks! :)
ı have tried all the transformation but still, it did not fixed the positive skew matter. what should we do?? please guide me
Hmm...have you also considered removing outliers (high Cook's D value) or cases with high leverage?
@@virginiawickline2609 is it normal to remove it. I am about to consider this variable as dependent one. It was behaving like poisson but with this behavior it doesn't respect poisson assumption either.
@@phd.eriolamariuscharlotade5634 Hi! You are reaching the limits of my knowledge. However, it is fairly common to remove cases with high Cook's D values so you are removing the influence of outliers. See, for example: towardsdatascience.com/identifying-outliers-in-linear-regression-cooks-distance-9e212e9136a