This is the best video I've ever seen. If there were more math teachers that explained this well a lot of children would love maths. Excellent explanation, thank u.
In my research, there are two groups: control and experimental. Both of these groups gave pre- and post-tests. Which test should I use? In the experimental group, there are 100 people,50 male, 50 female and in the control group, there are 100 people.50 male, 50 female, Could you please explain how to calculate the mean and standard deviation for this large sample? Thank you!”
Z-test is generally only used if you know the population variance (which is usually not the case). For a large sample size, the Z-test and t-test will result in about the same p-value.
Since we have a paired study design we should use a test that has been designed for this type of study, which is the paired t-test in this example. One assumption for the unpaired t-test is that the two groups are independent, which is not the case for the example in the video since the same persons are included in both "groups" (before and after).
The two distributions don't need to be normally distributed. That's a common misconception. Only the estimate of the mean needs to be normally distributed
Yes, but the underlying distribution that you sample from has to be normal, or close to normal, in order for the sample means to have a normal distribution if the sample size is small. I have a video that explains this: ruclips.net/video/NTq0BO6pbbg/видео.html
@@tilestats That's not true actually. It can be any distribution, as long as it has a defined standard deviation, and you are sampling a sufficient size.
@@tilestats Quote: "When we use an unpaired t-test, we usually assume that the observations in both groups are normally distributed, whereas the paired t-test only assumes that the differences follow a normal distribution." What this would mean is: Before ~ N(mean_before, sigma_before), After ~ N(mean_after, sigma_after) for unpaired, and Difference ~ N(mean_difference, sigma_difference) for paired. And that's not true. All we need is mean^hat_before ~ N(some_mean, some_sigma) for any of those distributions. Before, After, and Difference can be ~ Any(with_mean, with_sigma).
Jag undervisar bara på internationella program så jag har inget material på svenska. Men det skulle vara kul att göra några videos på svenska nån gång.
This is the best video I've ever seen. If there were more math teachers that explained this well a lot of children would love maths. Excellent explanation, thank u.
Thank you!
Thanks, now the difference between paired and unpaired T-test is so clear after this video.
you are the genious if statistics
Thank you for this video. Granted stats is my Achilles heal, so I did have to watch it twice before I understood it
You are a freaking unit man. Let's go
Very well explained! Great work :)
In my research, there are two groups: control and experimental. Both of these groups gave pre- and post-tests. Which test should I use? In the experimental group, there are 100 people,50 male, 50 female and in the control group, there are 100 people.50 male, 50 female, Could you please explain how to calculate the mean and standard deviation for this large sample? Thank you!”
You can use an unpaired t-test between exp group and control group based on the differences between pre and post test.
@@tilestats thanku so much, If my population is larger, do I still need to use a paired t-test over a z-test?
Z-test is generally only used if you know the population variance (which is usually not the case). For a large sample size, the Z-test and t-test will result in about the same p-value.
@@tilestats thanku so much.
Could you please elaborate why unpaired test is not used here ??
Since we have a paired study design we should use a test that has been designed for this type of study, which is the paired t-test in this example. One assumption for the unpaired t-test is that the two groups are independent, which is not the case for the example in the video since the same persons are included in both "groups" (before and after).
Thanks a lot! Excelent explanation!
Thank you!
The two distributions don't need to be normally distributed. That's a common misconception. Only the estimate of the mean needs to be normally distributed
Yes, but the underlying distribution that you sample from has to be normal, or close to normal, in order for the sample means to have a normal distribution if the sample size is small. I have a video that explains this: ruclips.net/video/NTq0BO6pbbg/видео.html
@@tilestats That's not true actually. It can be any distribution, as long as it has a defined standard deviation, and you are sampling a sufficient size.
That is exactly what I say in that video.
@@tilestats Quote: "When we use an unpaired t-test, we usually assume that the observations in both groups are normally distributed, whereas the paired t-test only assumes that the differences follow a normal distribution."
What this would mean is: Before ~ N(mean_before, sigma_before), After ~ N(mean_after, sigma_after) for unpaired, and Difference ~ N(mean_difference, sigma_difference) for paired.
And that's not true. All we need is mean^hat_before ~ N(some_mean, some_sigma) for any of those distributions.
Before, After, and Difference can be ~ Any(with_mean, with_sigma).
Yes
thanks bro helped a lot
Great!
Din svenska accent går inte att ta miste på. Vrf gör du inte videon på svenska?
Jag undervisar bara på internationella program så jag har inget material på svenska. Men det skulle vara kul att göra några videos på svenska nån gång.
@@tilestats Tack för ditt svar och lycka till med allt!