this is the 20th times I relearn this concept for the past 18 years. I like hw you focus on the concept not the calculation. Best Stat lecture I've seen so far !
The interpretation of this lecture is surprisingly thorough and easy to understand. When I read it carefully, everything makes sense, puzzle pieces together. It's so so so so much different from the gibberish I often heard in class - inconsistency, interchange concepts irresponsibly without notice, unidentified Subject, Object or Verb in the sentence :v I'm gonna watched it till the end, thank you Professor Marin.
Thank you Mike for explaining how S.E.M of mean is calculated in a previous video. It helped me understand the S.E. of difference of mean in this video. You are a great instructor. Thanks again for also explaining the necessary concepts.
At 13:00 why do we conclude that the brain injury group sleeps more if we did a two tailed test. Why not just conclude they sleep more or less (i.e differently) to the non-brain-injury group.
no, because in a confidence interval the t-value of t that gets you the level of confidence you want (eg the value of t that captures 95% of the distribution), while for the hypothesis test the t-stat is calculating how many SEs your estimate is from the hypothesized value. this video may help explain more... ua-cam.com/video/J-yMiTaai4c/v-deo.html
Wait, t-value=2.92, which affect the: CI = 0.7 +- (2.92)*0.24. => CI = -0.0008 to 1.4. How do I thing of it? The t-value is just a guide? Zero is in the CI? Should still I reject h0?
T value should be 2.04 as we calculate confidence interval for 95% that means 2 standard deviation, therefore z value should be 2 and similarly t value should be 2.04.
Yes, exactly. I left that of this video, to focus on the concepts and not the calculations. In a separate video before this one I explain the calculation of the SE..you can see that here if you want the details: ua-cam.com/video/ikS7itcmWZM/v-deo.html
Just in terms of nomenclature, I am not 100% sure what the difference between one sample, two sample and paired t-tests are. It seems that at the end, we do the same calculation, but perhaps how we come to the "t distribution" to be used in the calculation is slightly different. (mostly how we calculate sd it seems.) Is this correct or am I missing something? Perhaps a lecture just highlighting the difference and commonality might be useful in solidifying the concepts. Thanks! These lectures and matching R lectures are wonderful.
yes, they're pretty similar...here's the difference: The One Sample t-test: is for when you have one numeric variable and you will test about the mean. so you have the mean for one group. ex. is the mean blood pressure for a population greater than 120? The Paired t-test: is for when you want to compare the mean of two different groups, but the groups are paired/dependent/matched in some way. ex. you have people's blood pressure Before some treatment and then their bold pressure After a treatment. you want to compare the mean before to the mean after...and the before/after are paired...they are the same people. The Two Sample t-test: is for what you want to compare the mean of two different groups, and the groups are independent of one another. ex. you have one set of people and you give them treatment A and the measure their bold pressure. you have a different set of people (independent) and you give them treatment B and then measure their blood pressure. you want to compare the mean blood pressure for group A and for group B. so, they're all very much related. one sample t-test is for tests about one mean two sample t-test is for comparing means of 2 groups that are independent paired t-test is for comparing means of 2 groups that are paired/matched
If we already know the population mean, then why even do all this? I don't get it. Why even take a sample from a population whose characteristics are already known?
Dear Dr. Martin, I have been watching your chanel - it has great videos! Could you kindly redirect me if you have any videos on sample size calculations?
this is the 20th times I relearn this concept for the past 18 years. I like hw you focus on the concept not the calculation. Best Stat lecture I've seen so far !
The interpretation of this lecture is surprisingly thorough and easy to understand. When I read it carefully, everything makes sense, puzzle pieces together. It's so so so so much different from the gibberish I often heard in class - inconsistency, interchange concepts irresponsibly without notice, unidentified Subject, Object or Verb in the sentence :v
I'm gonna watched it till the end, thank you Professor Marin.
love this guy, he’s so easy to understand
Thank you Mike for explaining how S.E.M of mean is calculated in a previous video. It helped me understand the S.E. of difference of mean in this video. You are a great instructor. Thanks again for also explaining the necessary concepts.
At 13:00 why do we conclude that the brain injury group sleeps more if we did a two tailed test. Why not just conclude they sleep more or less (i.e differently) to the non-brain-injury group.
How can this man writing inversely?
All your tutorials are great! Cheers and salutations from Argentina!
Thanks! Glad you enjoyed them
Thank you so very much for providing this instruction, could not have made it through my research project without you.
Glad you found the videos useful @Paula!
great lecture...
Should not be t calculated in CI is same as t(stat) calculated in the hypothesis test?
no, because in a confidence interval the t-value of t that gets you the level of confidence you want (eg the value of t that captures 95% of the distribution), while for the hypothesis test the t-stat is calculating how many SEs your estimate is from the hypothesized value.
this video may help explain more... ua-cam.com/video/J-yMiTaai4c/v-deo.html
I made the same error
@@marinstatlectures but shouldn't we divide Standard error by root n, like we did everywhere previously?
please explain how did you get the p-value ?
Hi Marin.. U r awesome.. u make the concept crystal clear and you have a beautiful handwriting
thanks :)
How did u get .24?
He uses the squar(Var(mean(Yinj) - mean(Yno))), using the properties of variance you find the formula: squar((SD1^2)/n1 + (SD2^2)/n2)
Wait, t-value=2.92, which affect the: CI = 0.7 +- (2.92)*0.24. => CI = -0.0008 to 1.4. How do I thing of it? The t-value is just a guide? Zero is in the CI? Should still I reject h0?
T value should be 2.04 as we calculate confidence interval for 95% that means 2 standard deviation, therefore z value should be 2 and similarly t value should be 2.04.
No, t that you meant is t stat not the tvalue. So here from what he already explained that t val for 95% confidence was roughly 2. CMIIW
Hello Mike,
How did you find the standard error? By the pooled estimation or something like that?
Yes, exactly. I left that of this video, to focus on the concepts and not the calculations. In a separate video before this one I explain the calculation of the SE..you can see that here if you want the details: ua-cam.com/video/ikS7itcmWZM/v-deo.html
Just in terms of nomenclature, I am not 100% sure what the difference between one sample, two sample and paired t-tests are. It seems that at the end, we do the same calculation, but perhaps how we come to the "t distribution" to be used in the calculation is slightly different. (mostly how we calculate sd it seems.) Is this correct or am I missing something? Perhaps a lecture just highlighting the difference and commonality might be useful in solidifying the concepts. Thanks! These lectures and matching R lectures are wonderful.
yes, they're pretty similar...here's the difference:
The One Sample t-test: is for when you have one numeric variable and you will test about the mean. so you have the mean for one group. ex. is the mean blood pressure for a population greater than 120?
The Paired t-test: is for when you want to compare the mean of two different groups, but the groups are paired/dependent/matched in some way. ex. you have people's blood pressure Before some treatment and then their bold pressure After a treatment. you want to compare the mean before to the mean after...and the before/after are paired...they are the same people.
The Two Sample t-test: is for what you want to compare the mean of two different groups, and the groups are independent of one another. ex. you have one set of people and you give them treatment A and the measure their bold pressure. you have a different set of people (independent) and you give them treatment B and then measure their blood pressure. you want to compare the mean blood pressure for group A and for group B.
so, they're all very much related.
one sample t-test is for tests about one mean
two sample t-test is for comparing means of 2 groups that are independent
paired t-test is for comparing means of 2 groups that are paired/matched
@@marinstatlectures Thank you!
If we already know the population mean, then why even do all this? I don't get it. Why even take a sample from a population whose characteristics are already known?
Dear Dr. Martin,
I have been watching your chanel - it has great videos!
Could you kindly redirect me if you have any videos on sample size calculations?
wait how the frick do u write backwards so quickly