Amazing, I work at a hospital and i have to run some numbers from time to time to assess our quality improvement projects. this video is a life saver!!
This video and other videos in the series are godsend. I am a mid-career Aerospace Engineer, and have not needed heavy statistics but I have been wanting to study statistics for a long time. I am going to go through every video in this course. It's eye opening.
Great video, the only thing I noticed that you didnt explain the difference use of Wilcoxon and Man-whitney : the first is used for paired testing (Two dependent samples) the second is used for unpaired testing (Two independent samples)
Great video. A bit confused about the calculation for mean of the sampling distribution. The paper I read to set it up said it's n1*n2*0.5. Why does yours have the additional n1 outside the parentheses?
What if you want to do a two-tailed test on the data? If the hypothesis was "is there difference between fall and spring?" and not "is fall better/worse than spring?" How should you calculate Z?
Hey Brandon, PhD Student here looking to add some statistics to his research. First, love the videos. Stats is not in my wheelhouse so your videos have been extremely helpful clearly explaining everything and helping me work out what I need. I was wondering if you could either make a video or possibly provide a link that clearly explains how to work out and interpret the two varieties of the Mann Whitney test (e.g. shift location vs P(X
Great video, well explained. I just wonder why you calculate the z-value and not a t-statistic? The way I see it is that the samples are paired - as you measure improvement of the course comparing one semester to an other - so I would say the number of observations then is 25 and one would - strictly speaking - have to use the student t-statistic instead of the z-value? Or am I missing something?
Also what about the continuity correction? You are approximating a discrete distribution using the continious normal distribution, so you should in theory have to correct for this.
Just one question: Isn't the two samples dependent, since the professor is (for the most part) the same from one semester to the next? I thought that Rank-sum-tests were only for comparing two independent samples?
Brandon, Let's say you had 2 samples of n=20 and the total sums were the same (514, 761). However, now with the new numbers you will get a z = +2.81; thus not rejecting the Null, even though the Spring semester score is much higher. How is that this test depends only on the sample size, is that logical? Thanks Liran
Here's a weirder example: If you take 2 samples of n=20, and you assign the score of 5 to each line in Spring semester and 10 to Fall semester, you will get that the Null can not be rejected, meaning that there's no change. Am I wrong here or what?
If we are considering that the distributions of the data for Mann-Whitney-Wilcoxon is not normally distributed so why are we considering a normal distribution when calculating the p-value???
Why do you calculate the mean value (expected mean value )in this way? The original paper uses (n_1 * n_2) / 2 which is also equal to the sum of ranks / 2 namely (U_1 + U_2) / 2
Pretty good video, but the number of times you say "remember" throughout your videos makes it hard to listen to at times. The more things students have to "remember" from previous items, the harder things are to learn. Try to use less "remember" less and provide more clips/notes/animations on the slide to assist with the process. Thanks!
I must say finding this course was a blessing from the sky
I'm subject coordinator and I really loved it. Thank you for making this video.
Amazing, I work at a hospital and i have to run some numbers from time to time to assess our quality improvement projects. this video is a life saver!!
This video and other videos in the series are godsend. I am a mid-career Aerospace Engineer, and have not needed heavy statistics but I have been wanting to study statistics for a long time. I am going to go through every video in this course. It's eye opening.
Literally saved my masters 😂 THANK YOU!
Thanks for the video. It helped me a lot. Great explanation!
Great video, the only thing I noticed that you didnt explain the difference use of Wilcoxon and Man-whitney : the first is used for paired testing (Two dependent samples) the second is used for unpaired testing (Two independent samples)
PhD Student once again, THANK YOU!
So helpful. Thx for the knowledge
Great video. A bit confused about the calculation for mean of the sampling distribution. The paper I read to set it up said it's n1*n2*0.5. Why does yours have the additional n1 outside the parentheses?
What if you want to do a two-tailed test on the data? If the hypothesis was "is there difference between fall and spring?" and not "is fall better/worse than spring?"
How should you calculate Z?
Maybe just double that p-value since it's equally likely on either side of the normal distribution?
Truly helpful
I don't understand why 514 is being used and not 761.
Can someone please help.Thank you.
Amazing
Hey Brandon,
PhD Student here looking to add some statistics to his research.
First, love the videos. Stats is not in my wheelhouse so your videos have been extremely helpful clearly explaining everything and helping me work out what I need.
I was wondering if you could either make a video or possibly provide a link that clearly explains how to work out and interpret the two varieties of the Mann Whitney test (e.g. shift location vs P(X
Great video, well explained. I just wonder why you calculate the z-value and not a t-statistic? The way I see it is that the samples are paired - as you measure improvement of the course comparing one semester to an other - so I would say the number of observations then is 25 and one would - strictly speaking - have to use the student t-statistic instead of the z-value? Or am I missing something?
because it's not about the difference in scores of the same exact students
Also what about the continuity correction? You are approximating a discrete distribution using the continious normal distribution, so you should in theory have to correct for this.
Just one question: Isn't the two samples dependent, since the professor is (for the most part) the same from one semester to the next? I thought that Rank-sum-tests were only for comparing two independent samples?
Brandon,
Let's say you had 2 samples of n=20 and the total sums were the same (514, 761).
However, now with the new numbers you will get a z = +2.81; thus not rejecting the Null, even though the Spring semester score is much higher.
How is that this test depends only on the sample size, is that logical?
Thanks
Liran
Here's a weirder example:
If you take 2 samples of n=20, and you assign the score of 5 to each line in Spring semester and 10 to Fall semester, you will get that the Null can not be rejected, meaning that there's no change.
Am I wrong here or what?
Hi sir, is the denominator of standard Dev always 12?
If we are considering that the distributions of the data for Mann-Whitney-Wilcoxon is not normally distributed so why are we considering a normal distribution when calculating the p-value???
This is due to the sample size. If the sample size is large enough you can approximate, see 7:23
How u get that alpha value?
Someone just decided it because it's a very common practice to have an alpha of 0.05 or 0.01 for most tests.
Why do you calculate the mean value (expected mean value )in this way? The original paper uses (n_1 * n_2) / 2 which is also equal to the sum of ranks / 2 namely (U_1 + U_2) / 2
Pretty good video, but the number of times you say "remember" throughout your videos makes it hard to listen to at times. The more things students have to "remember" from previous items, the harder things are to learn. Try to use less "remember" less and provide more clips/notes/animations on the slide to assist with the process. Thanks!