Actually I studied it in 1978 and it doesn't seem too much has changed since then. But yeah there was no internet in those days and even by 1986 educational videos were expensive to make and distribute.
Since she's holding a power glove maybe this could be from the 80's version of RUclips, when you could order your crash course videotapes on tubes sent on the mail
I'm really glad this topic was given the room to breath over the last 3 episodes. Even in third year I come across people who have a complete misunderstanding of how p values work and what information they convey.
WisMicYal11 do you have any good material for me to study? Like a pdf containing some cheats of stats? Or some explaining stuff, like how to use p-value?
f(x | mu, sigma) = (1 / sqrt(2 * pi * sigma^2)) * e^-1*(((x-mu)^2)/(2*sigma^2)). There you go, a free PDF for you that is useful to do stats. Wanna learn about CDFs next?
It was a play on the abbreviation PDF, which in everything speach means the file format for documents that stands for Portable Document Format, but the same abbreviation in statistics means Probability Density Function, which are mathematical functions used to describe what a statistical distribution looks like. The math formula I gave you is the PDF and is the formula for the normal distribution. Were you to put in that formula in your graphical calculator and substitute 0 as mean and 1 as sigma and plot across a window of x values between -5 and 5, you'll see a standard normal distribution emerge. Like the statistical PDF, the CDF or Cumulative Distribution Function also describes a statistical distribution but as a sum of all probabilities below the x-position you're investigating.
Mistake on type I\ type II error visualizations? I think There is a mistake in the visualization of the error types on ~6:20; Moving the threshold ('cut-off-line') to the righter regions from the point of two of the distributions equality: while type I error is indeed the area from the threshold to the right and under the H_0 distribution, type II error should be to the left of the threshold and under the H_1 distribution, but it is not marked that way. (It is actually marked the area that is under both (minimum) of distributions, and to the right of the threshold) the same goes for the visualizations where the threshold is moved to the left (~6:30).
6:20 Type II error should not be bounded by null distribution, and Type I should not be bounded by alternative distribution. This animation is misleading because it illustrates a fixed error sum with respect to cutoff. In reality, setting the cutoff at the intersection of the distributions *minimizes* this sum.
I'm not fluent in Bayesian statistics, but as I understand it: the intuition is that credible intervals are the Bayesian version of confidence intervals that Frequentists use.
To the sound guy: I like the new sound effects. FWIW, it reminds me of a failed game that I tried playing for a while called "Godus." ...It also seems to be taking some tonal cues from the Endless series (e.g.: Endless Space)...
*_...so-in votation we'd avoid the midrange 'indistinguishable-from-a-drunk vote' entropic lobe but here we're asking whether the high and low ends of voting are also some kind of 'indistinguishable-from-silliness vote' (cf polling members for their likelihood of voting for, and separately, voting against; A test of 'claimancy')..._*
The statement about smoke alarms is insane. There is a massive cost to false alarms: they are so common as to turn "smoke detectors" into what they really are, toast detectors. All you do, when you hear the alarm, is shut it off and go back to your routine, because you don't care that someone is making toast. That is, the false alarms almost guarantee that the sound of the toast detector will be ignored when there actually is a fire.
The graphic is incorrect, type 1 error is the area under the null distribution to one side of the cutoff and the type 2 error is under the alternative distribution--neither is under both curves
i get that it takes time to shoot these videos; but wouldn't it be better if the upload rates were faster than once a week , maybe once a day or 2 days
I assume they don't want to overload people with too much information all at once and this method allows them to have many different series running simultaneously to draw in different kinds of learners
Keep in mind they're also using the same studio and staff to shoot and edit a whole bunch of other channel uploads as well. It may seem well within their capability to produce this series faster, but they've also got to ensure they keep to schedule for all those other shows (such as the other crash courses) they produce as well. All while also remaining within budget, which is probably why they stick to using this one studio and/or not hiring more staff.
yeah i get it since it is a crash course it would be better if the viewer is able to binge watch it; maybe just shoot the entire series and then upload all the videos at once
Then people would complain about the looong breaks in between different series. You'd binge-watch it in a few days (or even just one day!) and then have to wait several months for a new series to come out. The comments would be full of people asking when the next new series is coming out.
I wish this course was available in 1986. I'm finally understanding statistics. Thanks!
Teresa White looser oldman your generation sucks
@William Pollo, I wish your comment was not available in 2018.
Actually I studied it in 1978 and it doesn't seem too much has changed since then. But yeah there was no internet in those days and even by 1986 educational videos were expensive to make and distribute.
Since she's holding a power glove maybe this could be from the 80's version of RUclips, when you could order your crash course videotapes on tubes sent on the mail
Who are we?! Statisticians!! What do we want?!? Shinier cats!!!
Came for shiny cats comment, was not disappointed.
I'm really glad this topic was given the room to breath over the last 3 episodes. Even in third year I come across people who have a complete misunderstanding of how p values work and what information they convey.
I teach stats. I'll definitely send these videos to my students who struggle if they are interested.
WisMicYal11 do you have any good material for me to study? Like a pdf containing some cheats of stats? Or some explaining stuff, like how to use p-value?
f(x | mu, sigma) = (1 / sqrt(2 * pi * sigma^2)) * e^-1*(((x-mu)^2)/(2*sigma^2)). There you go, a free PDF for you that is useful to do stats. Wanna learn about CDFs next?
Teun van den Brand where is the pdf?
It was a play on the abbreviation PDF, which in everything speach means the file format for documents that stands for Portable Document Format, but the same abbreviation in statistics means Probability Density Function, which are mathematical functions used to describe what a statistical distribution looks like. The math formula I gave you is the PDF and is the formula for the normal distribution. Were you to put in that formula in your graphical calculator and substitute 0 as mean and 1 as sigma and plot across a window of x values between -5 and 5, you'll see a standard normal distribution emerge. Like the statistical PDF, the CDF or Cumulative Distribution Function also describes a statistical distribution but as a sum of all probabilities below the x-position you're investigating.
Teun van den Brand hahahahaha Nice! But it was too clever for me to understand hahahaha
Your course is amazing. Even though i took more than 3 statistics courses, i learn many things from you. Please keep up the great work!
7:10
if the null is true, you will fail to reject it 1-alpha of the time.
Yeah they did some serious mistake here
@@FootLettuce I was playing this sentence back and forth, again and again, lol
@@scarlettchen9943 Me too! I was so confused at first
Mistake on type I\ type II error visualizations?
I think There is a mistake in the visualization of the error types on ~6:20;
Moving the threshold ('cut-off-line') to the righter regions from the point of two of the distributions equality: while type I error is indeed the area from the threshold to the right and under the H_0 distribution,
type II error should be to the left of the threshold and under the H_1 distribution, but it is not marked that way.
(It is actually marked the area that is under both (minimum) of distributions, and to the right of the threshold)
the same goes for the visualizations where the threshold is moved to the left (~6:30).
With great power, comes great hypothesis testing, also, I’ve got the (statistical) power!
Thanks for the crash course! Great background on understanding A/B test results.
During the video I was like "what's the glove for?" and at the end I was laughing out loud hahahahahahahahahaha
That was pretty hard-core commitment to the bit, for a power-glove joke.
That Power Glove...
6:20 Type II error should not be bounded by null distribution, and Type I should not be bounded by alternative distribution. This animation is misleading because it illustrates a fixed error sum with respect to cutoff. In reality, setting the cutoff at the intersection of the distributions *minimizes* this sum.
7:13
"When the alpha is 0.05 that means that when the null is true, we'll correctly fail to reject [the null] 0.95 or 95% of the time"
finally i feel i understand something hope finish last part today
Excellent, Excellent Presentation!!! Please keep these good stuff coming!
I hope we talk about credible intervals soon. I'm still unclear how they differ from confidence intervals.
I'm not fluent in Bayesian statistics, but as I understand it: the intuition is that credible intervals are the Bayesian version of confidence intervals that Frequentists use.
Great video. Helped me a lot!
Thank you! Awesome visuals. Please keep them coming.
Even she couldn't help but crack up at saying 'WE WANT SHINIER CATS!'
Grate video! Leaning a lot about stat!
this is the single ep ill give to ppl if i can only get them still 1 set of 10mins at best
So now we are playing with power=)
Can we have a crash course Geology and Crash course Medicine
To the sound guy: I like the new sound effects. FWIW, it reminds me of a failed game that I tried playing for a while called "Godus." ...It also seems to be taking some tonal cues from the Endless series (e.g.: Endless Space)...
Excellent video
So early to learn stuff.
This is great, thank you!
4:42 it is misleading to make the cutoff line looks like a line of symmetry
yeah I was also confused, but then they started shifting it left and right which cleared it up a bit.
*_...so-in votation we'd avoid the midrange 'indistinguishable-from-a-drunk vote' entropic lobe but here we're asking whether the high and low ends of voting are also some kind of 'indistinguishable-from-silliness vote' (cf polling members for their likelihood of voting for, and separately, voting against; A test of 'claimancy')..._*
At 6:15 an onward, the shading is not correct. It should have been shaded above the overlap as well.
Hey There I Am From Bangladesh💜💜👌👌
We want...shinier cats? Is that a thing?
"It is a power. I want powers" - Pete Buttigieg
The statement about smoke alarms is insane. There is a massive cost to false alarms: they are so common as to turn "smoke detectors" into what they really are, toast detectors. All you do, when you hear the alarm, is shut it off and go back to your routine, because you don't care that someone is making toast. That is, the false alarms almost guarantee that the sound of the toast detector will be ignored when there actually is a fire.
Nothing is true. Everything is rejected.
Kids these days don't know about the powerglove
Ooh! Shiny!
It is over 😦😦😦😦😦😦😦😦😦😦😦😦😦
not even close...
The graphic is incorrect, type 1 error is the area under the null distribution to one side of the cutoff and the type 2 error is under the alternative distribution--neither is under both curves
Not sure what you're seeing, but their graphs are correct.
Yayyyyy!!!!!!
no idea what’s she saying
P-values? more like Power-values!
Can't we get this is layman language that is more remembrable
this is as layman's as you'll get. It does sometimes acquire a bit of pausing the video and pondering but that's just how ur supposed to learn
NULL
who's that babe in the tub full of flowers?
i get that it takes time to shoot these videos; but wouldn't it be better if the upload rates were faster than once a week , maybe once a day or 2 days
I assume they don't want to overload people with too much information all at once and this method allows them to have many different series running simultaneously to draw in different kinds of learners
Keep in mind they're also using the same studio and staff to shoot and edit a whole bunch of other channel uploads as well. It may seem well within their capability to produce this series faster, but they've also got to ensure they keep to schedule for all those other shows (such as the other crash courses) they produce as well. All while also remaining within budget, which is probably why they stick to using this one studio and/or not hiring more staff.
yeah i get it since it is a crash course it would be better if the viewer is able to binge watch it; maybe just shoot the entire series and then upload all the videos at once
Then people would complain about the looong breaks in between different series. You'd binge-watch it in a few days (or even just one day!) and then have to wait several months for a new series to come out. The comments would be full of people asking when the next new series is coming out.
Second
hurry up
So early to learn stuff.