Hypothesis Testing and P-values

[1rachellund]

please number your wonderful videos cheers



[krescent2]

why not using the T test?

[DesDemona1134]

Yay! I get it now! My instructor made is SO confusing. This was beautifully simple. Thank you, thank you, thank you!

[barrierloss]

Explained in a minute what my professor couldn't convey given hours daily to teach it

[beer94]

waait.... why did you reject? if theres less than .003 chance that the drug works, then doesnt that mean that the drug doesnt work??? so you fail to reject that the drug has no effect

[Riou2294]

@beer94 It's saying if the null is true, theres a .003 probablility that the rats have a response time as extreme as 1.05.

And the sample of rats that got DRUGGED has a MEAN of 1.05. So it is very likely that the drug affects the response time of the rats.

Therefore, the null hypothesis is REJECTED and the alternative is accepted. "Drug has an effect"

[Riou2294]

@beer94 it is NOT saying that theres less than .003 chance that the drug works.

[dragonfilth13]

@Riou2294 It is stating that the p=0.003 and therefore the null hypothesis should be rejected. if the Z score was more than 0.5, the alternate hypothesis should be rejected. So in affect. 1 in every 300 experiments, the Null hypothesis would be correct.

[bgoodwy2k]

this confused me more... I guess used to going to z table and comparing the z score to the probability... arrgh

[midgetman8888]

99.7 = 99.74?

[ubertuberdude]

saved my ass dude.

[ColdByrdz]

9:29 xD

[TheCrink15]

THANK YOU, NEEDED THIS FOR MY FINAL TOMORROW :D

[iamsopoor9]

10:00 GUYS! He's going to reject!

[YukiUchihaxX]

wait...0.003 <<< where did it come from?!

[midgetman8888]

@YukiUchihaxX if 1 = 100% then 0.3% is 0.003

[amann24]

Wish I had found these sooner, I struggled so bad with econometrics. You make things so easy to understand.

[biebergirl3194x]

so is the p value always just (1- the area from z score?)

[hatemzayed]

This is amazing Mr. Khan, no wonder why google give you 2 million dollars for your academy. Keep doing what you doing may God bless you for helping others.

[sanjaysinghiaf]

hi, if we do not have the population sigma, we do a t-test and thus use t-distribution as against z-test when we do have the population sigma.right? then how come u used the z-test?

[Swarm50]

@sanjaysinghiaf As your sample size "n" increases and becomes really large, the t-distribution becomes very similar and almost indistinguishable to the z-distribution. If you look at the Z and T tables you'll see this is true. It's only for small sample sizes, lets say under 50, that the t-distribution is really important.

[sanjaysinghiaf]

@Swarm50 . thx

[Saronaist]

I thank Jesus for you. You are the best :-)

[dijaro19]

i thought it was the sample mean minus the mean

[judyshin1220]

Are you really going to reject it? 10:01 Ha ha ha! ;)

[MaureenT1953]

I love Khan Academy and think Sal Khan is a hero. If only more teachers could explain concepts the way he can! Maybe he should do a series on 'how to be a better teacher' :-)

[capsunful]

I love you khanacademy!

[CollMomo1]

@NateC0le thats exactly my situation, ive been looking through youtube to understand all of these concepts, stats should be taken in class, not online, without atleast videos !!

[gwoppinallday]

Why does Sal estimate the standard deviation of the sampling distribution as 0.05 seconds? Is it not already given as 0.5 seconds.

[cpowel2]

@gwoppinallday

Because we need to derive the sample standard deviation for the problem which by definition is s/sqrt(n) which in this case = 0.5/10 = 0.05

[UnstoppableBird]

woww thank you dude!! 

[CMBingle]

SOOOOOO helpful. Thank you for speaking in plain words. So much easier than my long excessively verbose text book

[bgmakiinmcqu33n]

THANK YOU SO MUCH

[ssips720]

the magical math stoner... yes!

[totallykeet]

anyone think its bizarre how one guy is teaching every subject on earth

[medardoangelsilva]

As others have noted, isn't the formula: Z= sample mean minus population mean divided by population standard deviation divided by sample size squared. 1.05-1.20/ .05 which is -3. I think this has been noted, but should be changed in the video, because as it is the same result, I don't think this would be in a one-tiered test. I think this mistake could be a good way to better explain p-value.

[SaruwatariKaitoKun]

ur the best teacher ive ever had! Thank you Thank you Thank you

[kiwituatara]

I looked up the Z table for 3, its 0.9987. so 1 - 0.9987 = 0.0013?

[ternaldo]

I'm watching this video and I think "Why am I paying tuition to be schooled by youtube?" I would much rather sit in a lecture hall and have your videos play instead of getting frustrated over some boring monotone prof

[TheWingsofJustice]

@ternaldo I know! I wish my profs would just play Khan videos and shut up ...

[sfurtwengler]

Help! I thought z = M – μ / σ.

Isn’t the demo indicating z = μ – M / σ ? z = 1.2 – 1.05? What am I not getting?

[mmanch01]

This lesson does a good job at conveying the idea that if the probability of values beyond our sample mean is extremely small, then the claim that mu is as stated in the null hypothesis should be rejected. However, since sigma is not given, you should have conducted the test with a student t-distribution, not the normal distribution. Also, some discussion about 'confidence level' is in place - for example, if we demanded a confidence level of 99.9%, your results do not merit rejection of H_0.

[b3n5p34km4n]

@mmanch01 if n > 30 we need not replace the Gaussian with a t.

[sunflowerfeet]

Hey. Wouldn't you subtract the population mean from the sample mean, not the sample mean from the population mean? It makes your answer -3, which doesn't really matter in a two-tailed distribution, but would be important in a one tailed, right?

[Carebearbull]

Hypothesisisis lol. 

[2013CHUCK]

i so damn like ur interpretation of stats , it's better than my lucture which i have studying now , i feel like pian in my ass~~~~~~

[Elena25377]

Can you explain please the p value and error type 1 , error type 2, and power for the difference in means between two population means when all we know is the sample means and sample sizes?

[shivkalra1]

Hi, why didn't u use P(z < -3) instead of two tailed.

[poopdude]

Writing down what u actually means definitely helped me understand what claim I'm actually rejecting and what a low p-value means. Thank you! :D

[yasthegreatest1]

At 2:45. approximately you state that if the probability of it being not equal to the 1.2 is very small then that means H null is not true.. Why is that? Wouldnt that make it true since the probability of it being H1 is very small?

[usaswim323]

OH MY GOD I LOVE YOU Totally understand this now!

[miniprizes]

Thanks alot. I was stuck on my stats homework as to what a Ho and Halternative hypothesis were. Thanks!

[V2PRC]

I don't understand why you choose P(X<x bar), from my point of view P(X = x bar + dx) would be more reasonable.

[crazypankaj4u]

Hi Wat I didnt understood is U divided smaple mean by sq root of sample size assumed that as Population mean. By this defination if I have a extremely Large sample size the the SD of Smple and Population should be approx same. But with this formula and logic if I increase sample size I will Increase diffence bet sample and Population which I thought shud be other way round. Please explain this :)

[89Fad]

as we dont know the variance of the population we should use t-stats ! not z

[BITpinched]

@89Fad z-stat can also be used as long as the sample size is bigger than 30 (n>30). but of course the result from z-stat is only APPROXIMATE value whereby t-stat will give a more accurate result.

[braden873]

thanks you've saved my assignment!

[rrayoflight]

Tanx

[juanitoaustralia]

Thanks Sal!! This is what we need.

[norwayte]

Wonderful.