Thank you for providing this content. You have been quite helpful, and I believe this is not only for me. ●So, did I get correctly that if an A-D value is higher, it corresponds to a lower p-value and hence a higher risk of rejecting Ho that assumes the distribution comes from a specifed distribution (i.e., normal distribution) when the p-value is less than the significance level? ●Is it accurate to argue that the smaller the A-D value, the more likely your distribution is to be normal because the p-value will be larger (above the significance level), allowing Ho to assume that the distribution is from the specified distribution? ● The higher A-D value, the lower p-value and the higher chance of rejecting Ho? ●Otherwise, how should the A-D value be used in respect to the p-value? Thank you in advance.
Hello, Yes, you are correct that a higher AD value corresponds to a lower p-value and a higher risk of rejecting H0. And yes, the inverse of that would also be true: a smaller AD value corresponds to higher p-value and thus failing to reject the null hypothesis. - IMPORTANT - However, I would not recommend using the AD value to make your conclusions about normality because the p-value calculation also takes the sample size into account. This is one of the main reasons we use the p-value. It standardizes the criteria across sample sizes. Another important reason that we use p-values is because it is directly linked to the risk of Type I Error, this means that by specifying our alpha (generally 0.05) we are actually specifying the acceptable risk of Type I Error at our specific sample size. - CONCLUSION - The p-value is a standardized statistic that allows us to control Type I Error and to ensure that sample size effects are properly controlled. The AD test statistic is still important but should not be used by itself to make conclusions.
Hi, The Kuiper Test and Kolmogorov-Smirnov Test are both very similar. They both use the maximum discrepancies from the CDF to calculate a test statistic. The difference between the two is that the Kolmogorov-Smirnov Test Statistic is the overall maximum discrepancy MAX(D+, D-) whereas the Kuiper Test Statistic is the sum of the maximum discrepancies above and below the CDF SUM(D+, D-).
thankyou so much! i am from Indonesia and currently studied statistics in college, this video really helps me out, definitely would recommend you to my friends!
That is great to hear! If you are interested in Normality testing, especially if you're studying statistics, you should check out this video: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-WKyAYsqTSCM.html It might change how you see the AD Test. Thanks, Lucius LaFromboise
I used to stress about managing normality in excel, but this video provided a fresh perspective. Much respect to the creator for simplifying it! Liked and Subbed!
'QE.CAPABILITYANALYSIS' is always returning #VALUE! error, even though all data is as per the video. Do you have sample excel file from the demo somewheare on the cloud? Can you share the link?