Hello ! Thank you a lot for you videos ! But I have a question: since everyone is correlated to some extent, shouldn't we use this formulation to calculate the 'effective size' for each sample (M.yellow, etc.)?
And yes everyone should buy your book !!! It's an amazing book, and the clearest statistics courses I have seen!!! !! Please don't stop teaching like you do, we need you so much !!
Such a great series! Really helps make statistics approachable to us! I can't wait for a Central limit therm, delta method, strong law of large numbers or perhaps slutsky's theorem. Those are all topics which I found really challenging when I first started my studies.
Thank you so much for another helpful video. When discussing "effect size" and "effective sample size," we discuss two totally different concepts, right?
dear Josh, I will be thankful if you make a video about the effect size and cohen's d. these topics are intangible and I didn't find a good video for them on the net. thanks in advance.
Thanks for these awesome sets of lectures, sir. I had a question. does the formula work only for positive correlations? I mean what if the correlation between twins is negative so that they zeroed the denominator, or for example, make the effective size negative?
To be honest, all I know about the actual formula is that it's a little more complicated than what I presented (this video was simply to present the main ideas of the concepts) and presumably can handle negative correlations correctly.
Amazing as always, but here when you mention correlation are you referring to pearson correlation coeff? Is it possible for them to be non-linearly correlated (I cannot think of a situation)?
Great question! I'm not an expert on this one, but here's a guess: One way you could do it is measure expression from a lot of genes (perhaps all of them) and then calculate correlations based on that. Since there is going to be a fair amount of correlation between everyone (blue, orange and green - since they all are dudes and do dude things) you could scale the correlation based on how correlated blue is to orange and green.
@@statquest This sounds very important for any research and yet seems to be a methodological detail, i'm confused. Could you give us some examples? Thanks
@@statquest like other formula sample size, there is rule that margin error about 5 % - 10 % but not in 7%. For example like slovin formula is have margin error 5% and 10%. So that, i wanna ask can we use margin error 7% and is there any references that i can refers it? Thank you, sir
@@salsabillashafaadzra8109 I see. Unfortunately I don't know of any references. That doesn't mean they don't exist, it just means I don't know about them. bummer! :(
I think I just used to the built in mic on my laptop for this one. However, if you want details on my setup, see: th-cam.com/video/crLXJG-EAhk/w-d-xo.html
I have a couple questions about this video. Is there an explanation for the effective sample size equation you show? Also, why would the effective sample size equation not be linear with correlation? If you have a sample size of 2, and a correlation of 0.5, why would the effective sample size be 1.33 and not 1.5?
Hi, Here you calculated the effective sample size for persons belonging to the same set of population( blue ones), However there must be some correlation between the orange and blue ones or blue and green ones or orange and green ones or among all three of them .Aren't we supposed to calculate effective sample size for them too?
If all samples have the same amount of correlation, then it all washes out and we don't have to worry about things. However, if some samples are more correlated than others, then we need to take that into account, so that's what's going on here. In your own study, you need to figure out if there is a uniform amount of correlation among your samples or not. If not, then you need to adjust for it.
@@statquest Hi, how do you calculate this correlation? Using genetic correlation? Like for the identical twins, the genetic correlation = 1, then sample size = effective sample size. Then that means it won't raise the power by adding an identical twin dude into the sample, right? thanks.
Nice explanation. I have a question. I want to find the effects of Vitamin A on the methylation status of some specific genes in children. Now how to calculate the sample size?
Good question. You need an estimate of the effect size and the variation in the data. Then you can do a power analysis. Perhaps there are similar studies that you can look at for ideas.
BAM!! I Really like your lectures . They are super professional. Can you do on LSTM , Long range dependence, and other types of distributions like Pareto Distributions ....??. Thanks.
Hello, Joshua, I have an experiment on an F1 mapping population, I work with table grape, so I take 36 individuals of my population to measurement firmness 18 have a good firmness and the other not, actually are the bad ones.... so as you mentioned in this video do I have to apply an effective sample size? (Is it possible that you give your email?) thanks
Support StatQuest by buying my book The StatQuest Illustrated Guide to Machine Learning or a Study Guide or Merch!!! statquest.org/statquest-store/
Hello ! Thank you a lot for you videos !
But I have a question: since everyone is correlated to some extent, shouldn't we use this formulation to calculate the 'effective size' for each sample (M.yellow, etc.)?
More clearly: why this limitation, this threshold, with twins? Isn't it a big bias ?
And yes everyone should buy your book !!! It's an amazing book, and the clearest statistics courses I have seen!!!
!! Please don't stop teaching like you do, we need you so much !!
@@Also_sprach_Zarathustra. What time point in the video, minutes and seconds, are you asking about?
Thank you so much for explaining a student who got low marks in Statistics I'm feeling happy that my concepts are clear now. Can't thank you enough.
Thanks!
1:10 interesting, tree blood. This was a nice and sweet statquest.
thanks!
Such a great series! Really helps make statistics approachable to us! I can't wait for a Central limit therm, delta method, strong law of large numbers or perhaps slutsky's theorem. Those are all topics which I found really challenging when I first started my studies.
Thank you so much for another helpful video. When discussing "effect size" and "effective sample size," we discuss two totally different concepts, right?
Yes! Effect size is a measure of how big (or small) a difference there is between two groups.
dear Josh, I will be thankful if you make a video about the effect size and cohen's d. these topics are intangible and I didn't find a good video for them on the net. thanks in advance.
I believe I talk about these things in my video on power: th-cam.com/video/VX_M3tIyiYk/w-d-xo.html
Thanks for these awesome sets of lectures, sir.
I had a question. does the formula work only for positive correlations? I mean what if the correlation between twins is negative so that they zeroed the denominator, or for example, make the effective size negative?
To be honest, all I know about the actual formula is that it's a little more complicated than what I presented (this video was simply to present the main ideas of the concepts) and presumably can handle negative correlations correctly.
@@statquest Cool! Thank you for the clarification, sir :-)
Amazing as always, but here when you mention correlation are you referring to pearson correlation coeff? Is it possible for them to be non-linearly correlated (I cannot think of a situation)?
I believe people usually use pearson's correlation coefficient. I can't imagine the correlation being non-linear, but you can always check for that.
Can you please make a video about effect size or randome effect model?
Is effect size and effective sample size the same thing ?
No. Effect size is related to how different two groups are.
If you're a mouse geneticist, you can think of blue dudes as a specific strain of mouse. DOUBLE BAMM!!!
Yes! :)
how do we calculate correlation between let's say, 5 blue dudes?
Great question! I'm not an expert on this one, but here's a guess: One way you could do it is measure expression from a lot of genes (perhaps all of them) and then calculate correlations based on that. Since there is going to be a fair amount of correlation between everyone (blue, orange and green - since they all are dudes and do dude things) you could scale the correlation based on how correlated blue is to orange and green.
@@statquest This sounds very important for any research and yet seems to be a methodological detail, i'm confused. Could you give us some examples? Thanks
Hi sir, can we use margin error 7% and is there any journal that refers it? Thank you
I'm not sure what you mean by your question. Can you elaborate on it or give more context?
@@statquest like other formula sample size, there is rule that margin error about 5 % - 10 % but not in 7%. For example like slovin formula is have margin error 5% and 10%. So that, i wanna ask can we use margin error 7% and is there any references that i can refers it? Thank you, sir
@@salsabillashafaadzra8109 I see. Unfortunately I don't know of any references. That doesn't mean they don't exist, it just means I don't know about them. bummer! :(
Clearly explained. I have something to request. Would you like to create a video about law of large numbers and order statistics?
I'll keep that in mind.
Which mic do you use?
I think I just used to the built in mic on my laptop for this one. However, if you want details on my setup, see: th-cam.com/video/crLXJG-EAhk/w-d-xo.html
I have a couple questions about this video. Is there an explanation for the effective sample size equation you show? Also, why would the effective sample size equation not be linear with correlation? If you have a sample size of 2, and a correlation of 0.5, why would the effective sample size be 1.33 and not 1.5?
For more details on the formula for effective sample size, check out the wikipedia article: en.wikipedia.org/wiki/Effective_sample_size
Hi, Here you calculated the effective sample size for persons belonging to the same set of population( blue ones), However there must be some correlation between the orange and blue ones or blue and green ones or orange and green ones or among all three of them .Aren't we supposed to calculate effective sample size for them too?
If all samples have the same amount of correlation, then it all washes out and we don't have to worry about things. However, if some samples are more correlated than others, then we need to take that into account, so that's what's going on here. In your own study, you need to figure out if there is a uniform amount of correlation among your samples or not. If not, then you need to adjust for it.
@@statquest Hi, how do you calculate this correlation? Using genetic correlation? Like for the identical twins, the genetic correlation = 1, then sample size = effective sample size. Then that means it won't raise the power by adding an identical twin dude into the sample, right? thanks.
Could u please do one on latent variable?
I'll add that to the to-do list.
thumbs up
Nice explanation. I have a question. I want to find the effects of Vitamin A on the methylation status of some specific genes in children. Now how to calculate the sample size?
Good question. You need an estimate of the effect size and the variation in the data. Then you can do a power analysis. Perhaps there are similar studies that you can look at for ideas.
BAM!! I Really like your lectures . They are super professional. Can you do on LSTM , Long range dependence, and other types of distributions like Pareto Distributions ....??. Thanks.
Cool, I have no reason to know this info, but it is really interesting.
bam! :)
I love one man and that man is Josh Starmer
Thanks! :)
These videos are awesome! Is this the same concept as ICC for cluster-RCT analysis? Thanks :)
Maybe. Unfortunately, I'm not familiar with those terms.
great explanation sir thank you
Thanks!
what a clear explanation
Thank you!
Very interesting! Thanks for this!
Thanks! :)
Well! Everything is clearly explained 😅
bam! :)
If you're a quality engineer, you can think of technical replicates as repeatability and reproducibility samples. TRIPLE BAMMM!
Yes! :)
love the series!!
Damn that intro goes hard
:)
Hello, Joshua, I have an experiment on an F1 mapping population, I work with table grape, so I take 36 individuals of my population to measurement firmness 18 have a good firmness and the other not, actually are the bad ones.... so as you mentioned in this video do I have to apply an effective sample size? (Is it possible that you give your email?) thanks
I don't know
You must know how to do magics. Otherwise how could you make things so easy!!
Thank you! :)
im high watching this on freewill and not algorithm recommendation. i forgot why
dang!
Thanks a ton Sir
:)
Thank you!
Thanks!
Neat!
:)
how would you add a fraction of a dude to a sample though..?
Ha! I guess you need to round up! :)
Duh hahaha .