YT YouTube2Text

YouTube2Text → Transcripts → YouTube transcript (VFMcGdWp0MQ)

YouTube transcript (VFMcGdWp0MQ) — Transcript

3,657 words · 531 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:07[Music]
0:17hello thank you for watching and welcome
0:19to the next video in my series on basic
0:21statistics now as usual a few things
0:24before we get started number one if
0:27you're watching this video because you
0:28were struggling in a class right now I
0:30want you to stay positive and keep your
0:32head up if you're watching this it means
0:34you've accomplished quite a bit already
0:36you're very smart and talented and you
0:38may have just hit a temporary rough
0:40patch now I know with the right amount
0:42of hard work practice and patience you
0:46can get through it I have faith in you
0:48many other people around you have faith
0:50in you so so should you number two
0:54please feel free to follow me here on
0:55YouTube on Twitter on Google+ or on
0:59LinkedIn
1:00that way when I upload a new video you
1:02know about it and it's always nice for
1:05me to connect with people who watch my
1:06videos online wherever in the world you
1:09may happen to be now on the topic of the
1:12video if you like it please give it a
1:15thumbs up share it with classmates or
1:17colleagues or put it on a playlist
1:19because that does encourage me to keep
1:21making them for you on the flip side if
1:23you think there is something I can do
1:25better please leave a constructive
1:27comment below the video and I will try
1:29to take this idea is into account when
1:31to make new ones for you and finally
1:34just keep in mind that these videos are
1:36meant for individuals who are relatively
1:38new to snats so I'm just going over
1:41basic concepts and I will be doing so in
1:44a very slow deliberate manner not only
1:48do I want you to know what's going on
1:50but also why and how to apply it so all
1:53that being said let's go ahead and get
1:57started so this video is the next in our
2:00series on hypothesis formulation and
2:03eventually hypothesis testing and like
2:06the last videos it's sometimes a concept
2:09that can give intro stat students
2:11headaches it can drive them crazy and
2:14the reason is that this concept runs
2:17counter to how we usually think about
2:19our everyday lives and of course this is
2:22the concept of type one and type two
2:25errors in hypothesis formulation and
2:28testing now remember in previous videos
2:31we talked about what hypotheses are we
2:33talked about how to set them up pull
2:35data out of problems and then set up our
2:37null hypothesis and our alternative
2:40hypothesis now remember the null
2:42hypothesis is what we assume to be true
2:46it's the status quo it's the information
2:48we are given or we again we assume to be
2:50the case the alternative is where we go
2:54next if we have to reject the null
2:57hypothesis so if we have to reject our
3:00assumption then we proceed on to the
3:03alternative now one thing we know by
3:05this point is that stats is never
3:08certain it's never 100% perfect so the
3:12way this usually works as far as
3:14hypothesis go we set up our hypothesis
3:17we go out and collect a sample we do
3:19some analysis on that sample and then we
3:22reach some conclusion about our
3:26hypothesis so the question is does our
3:29conclusion
3:30from our
3:31analysis match the actual state of
3:36reality so remember we're basing that
3:38conclusion off some sample we took and
3:41then we either reject or fail to reject
3:44our null hypothesis but the question is
3:48does our conclusion match the actual
3:51state of reality now we know it's not
3:55going to happen 100% of the time and
3:58that in a nut check
4:00is type one and type two error does our
4:04conclusion in our hypothesis test due to
4:06our analysis match the actual state of
4:11reality now this is not just some
4:13academic distinction here in real life
4:17in real studies in real research the
4:20type one and type two errors can have
4:23literally life or death consequences and
4:27I'm not OV exaggerating so in the
4:29examples I use in this video I'm going
4:31to show you that as starkly as I can now
4:34the way I'm going to proceed is to show
4:36you a couple of non-statistical real
4:39world examples that exhibit the idea of
4:42type one and type two error and then we
4:45will proceed into more formal
4:47statistical ways of putting it so all
4:50that being said let's go ahead and look
4:52at our first nonstatistical real world
4:57example now I call this the fire alarm
5:00hypothesis and unfortunately it's
5:02something a lot of us have to deal with
5:04at some point in our lives or in
5:06something similar you know similar
5:08situation and it goes something like
5:09this let's say you're going down the
5:12hallway at work or school sort of doing
5:14your own thing everything is normal as
5:17usual but you encounter a sudden smell
5:21of smoke like something's
5:24burning now you know that may mean a
5:27serious fire is taking place in in the
5:30building or it could be nothing serious
5:33maybe someone burned popcorn or
5:35something else in the microwave in the
5:37kitchen or cafeteria or wherever else
5:39you might be but the question is what do
5:41you do next you have to make some
5:43decision on what you're going to
5:47do so let's take a look at what might
5:50happen if you think the Smoky smell is
5:52nothing serious you may decide that your
5:54assumption that everything is normal is
5:57correct and therefore you will not pull
6:01the smoke alarm or the fire alarm so
6:03maybe someone burned something in the
6:05kitchen of course that's annoying but
6:07it's not a serious fire so you would
6:10just proceed on like everything was okay
6:13and of course you will not pull the fire
6:16alarm for that now if you think the
6:18Smoky smell is due to a serious fire you
6:22may reject your assumption that
6:24everything is normal and then you will
6:27pull the fire alarm so you have to reach
6:30some conclusion based on the evidence
6:33around you about the state of reality
6:37it's either a not serious fire or it's a
6:41very serious fire and then you have to
6:43decide what you're going to
6:47do so let's look at what might go wrong
6:49in this situation so of course you smell
6:51the smoke and you think to yourself This
6:54is not normal so you reject the
6:58assumption that everything is okay you
7:01reject your null
7:03hypothesis therefore you go ahead and
7:06pull the fire alarm so you rejected your
7:09null hypothesis that everything is
7:11normal and then you went ahead and
7:13pulled the fire alarm so the building is
7:15evacuated and the fire department
7:17arrives to
7:19investigate now after the investigation
7:22it is determined that there was no
7:24serious fire you quote falsely pull
7:30the fire
7:31alarm now here's the way we interpret
7:34that when you rejected your assumption
7:37that everything was okay so you looked
7:40at the situation you said this is not
7:41normal I'm going to reject my assumption
7:43that everything is normal and then you
7:46proceeded so when you rejected your
7:48assumption that everything was okay when
7:50it really was okay because there was no
7:54fire you committed type one error or in
7:58this case a false alarm so let's talk
8:02about type 1 error more formally so type
8:041 error is rejection of the Assumption
8:08or rejection of the null hypothesis when
8:12it should not have been
8:14rejected so in the Fire case you
8:18rejected your assumption that everything
8:20was normal but the reality was
8:23everything was okay so type 1 error is
8:26rejecting the N hypothesis when it
8:28should not be rejected
8:30you can think of it as incorrectly
8:32rejecting the null
8:34hypothesis in this case a false
8:38alarm and again this is a real world
8:40situation but is a perfect example of
8:43what type one error is rejecting the
8:46null when it should not have been
8:52rejected now let's look at what else
8:54might go wrong let's say you smell smoke
8:57and you think oh it's probably someone
8:59who burns their lunge in the microwave
9:02no big deal so you go ahead and just
9:04proceed along your way therefore you did
9:08not reject your assumption that
9:10everything is okay you upheld your null
9:13hypothesis you just sort of concluded
9:16it's not nothing serious so you or not
9:18something serious so you went ahead and
9:19just went on your way but let's say
9:23there is indeed a serious fire now no
9:26one is injured luckily but the entire
9:29building does burn to the
9:31ground so when you failed to reject your
9:35assumption that everything is okay when
9:37it really was not okay you committed
9:42type two
9:43error so let's formally Define this and
9:46then we'll compare the two so type two
9:48error failure to reject the Assumption
9:51or failure to reject the null hypothesis
9:55when it should have been
9:57rejected so incorrect ly not rejecting
10:01the null
10:02hypothesis so think about type one type
10:06one is when we reject the null
10:10hypothesis
10:12incorrectly type two is when we fail to
10:17reject the null hypothesis
10:20incorrectly so in the first case we pull
10:23the fire alarm when there was no fire in
10:26the second case we did not pull the fire
10:29alarm even though there was a fire so
10:33again we'll do some other examples but
10:35this sort of sets the stage for
10:37basically what type one and type two
10:40error as it relates to our conclusion we
10:43reach and the actual state of
10:51reality now we can set this up as a
10:53little hypothesis so our null hypothesis
10:55was the smoke is annoying but not
10:58serious everything is okay as usual
11:00there's no serious fire the alternative
11:04would be that smoke is evidence of a
11:06serious fire everything is not okay as
11:09usual there is a serious fire now we can
11:13set up a little hypothesis table here so
11:16on the left hand side we have the
11:18conclusion we reach and then on the top
11:23we have the actual condition or the
11:25actual state of reality so let go ahead
11:29and fill this in with some
11:32information now in the conclusion side
11:34we have we do not reject the null
11:37hypothesis so no serious fire or we
11:42reject our null hypothesis that
11:44everything is okay and there is some
11:47serious fire so those are the two
11:49conclusions we could reach in this
11:51situation now the actual condition of
11:55reality is there is no serious fire or
11:58there is is a serious
12:02fire now we notice that there are two
12:05correct conclusions here so if we do not
12:08reject the null hypothesis and there's
12:11no serious fire and the state of reality
12:14is that there is no serious fire that is
12:17a correct
12:19conclusion now if we reject the null
12:21hypothesis that everything is normal and
12:23there is a serious Fire based on our
12:26evidence and reality is that there is a
12:29serious fire that is also a correct
12:34conclusion so you can see are two
12:36correct conclusions there now the
12:38question is where is the type one and
12:41type two error in this
12:44chart well they go here so let's look at
12:47the type one error situation first so in
12:51the type 1 error we reach a conclusion
12:54where we reject the null hypothesis so
12:57in this case we re reject the assumption
13:00that everything is normal and we think
13:03there is a serious fire going on but the
13:06fire department arrives and they
13:09determine the actual condition is that
13:12there is no serious fire so therefore
13:15our conclusion does not match the actual
13:19reality now let's look at type two error
13:23so in type two error our conclusion was
13:26that we did not reject our null
13:28hypothesis and therefore we don't think
13:30there's a serious fire going on but the
13:33reality is there is a serious fire going
13:36on so we have a disconnect between our
13:38conclusion and the actual state of
13:41things so you can see that we have two
13:43correct conclusions and two errors type
13:46one error and type two
13:50error now in general the real world
13:53consequences of a type 2 error are much
13:56greater in this case a type two error
13:59error May mean the loss of property or
14:02even lives so think about this what
14:05happens if we commit a type one error in
14:07this case well we think there's a really
14:11serious fire so we pull the fire alarm
14:14that gets everyone out of the building
14:16and it brings the fire department to the
14:19location but it turns out there's no
14:22serious
14:23fire well pulling the fire alarm is
14:26going to be a very big inconvenience to
14:28everyone in the building but we pulled
14:32it we pulled the fire alarm because we'd
14:34rather be safe than sorry right so the
14:37minor inconvenience of the fire drill is
14:41no match of the consequences of type two
14:44error so look at type two we don't
14:47reject our null hypothesis we think
14:50there's no serious fire going on when in
14:52fact there is a serious fire so we don't
14:56pull the fire alarm what can the what
14:59can the consequences be here well if no
15:02one catches it in time the building
15:05could continue to burn because there is
15:07a serious fire and it could mean the
15:09loss of course property and lives so
15:12while the type one error in this case is
15:15a real world inconvenience the type two
15:18error can cost property and lives and
15:21that pattern tends to repeat itself
15:25throughout these examples
15:30so let's look at another example and I
15:32call this the andon hypothesis now
15:35you're probably thinking what is that
15:37well here's what it is in a
15:39manufacturing system such as the Toyota
15:41production system so Toyota the car
15:44company and other lean systems line
15:47workers have the power and indeed the
15:50responsibility to stop the entire
15:52production line when they notice a
15:55serious problem and this is done by
15:58pulling what's called an andon cord so
16:01you can see in these pictures all along
16:03the production line there's a cord like
16:05you might see on a bus or even a train
16:07and if they notice something is
16:09seriously wrong the worker is supposed
16:12to is actually required in this
16:14environment to pull it that will stop
16:17the entire production line dead in its
16:20tracks until that situation is
16:26resolved now the assumption is that the
16:28production line is running correctly
16:31this is the minute-by-minute null
16:32hypothesis so as things run the
16:35assumption is everything is running
16:37smoothly everything is running correctly
16:39now when a worker notices a defect he or
16:42she has a choice to make should they
16:45pull the chord or
16:49not if the endon chord is pulled the
16:53worker is rejecting the assumption that
16:55the process is error free they are
16:59rejecting the null hypothesis they are
17:01rejecting the assumption that everything
17:03is okay if the core is not pulled then
17:08the worker is not rejecting the
17:10assumption that the process is error
17:12free thus maintaining the null
17:14hypothesis and the production line
17:16continues to run so it's all about
17:19rejecting or not rejecting the
17:21assumption that everything is
17:26okay now what if the chord is P which is
17:30a rejection of the null hypothesis but
17:33nothing is actually wrong with the
17:36production process again this is sort of
17:39a false alarm
17:41situation in this case the worker quote
17:44committed a type one error they reject
17:48the null hypothesis incorrectly they
17:51rejected the null hypothesis that
17:53everything was okay
17:56incorrectly so type one error again is a
17:59wrongful rejection of the null
18:01hypothesis when it should not be
18:03rejected now something about this in the
18:06Toyota production system and in similar
18:09Lean Manufacturing Systems this is
18:12actually encouraged this is perfectly
18:16okay so if the worker thinks there is
18:19something wrong they are supposed to
18:21they are required to reject their null
18:25hypothesis reject the assumption that
18:27everything is going okay pull the cord
18:30and stop the production line they are
18:32not supposed to worry about whether or
18:35not they are correct or incorrect and
18:39that goes to the idea of the difference
18:41in the real world consequences between
18:43type 1 and type two error which we'll
18:45talk about here in a minute but actually
18:48this type one error is okay in this type
18:52of production
18:53[Music]
18:55system now what if the cord is not
18:57pulled thus supporting the null
19:00hypothesis but something actually is
19:03wrong with the production process so
19:06often times a worker or sometimes a
19:08worker might be afraid to pull that cord
19:12because they don't want to be blamed for
19:14stopping production and this is
19:16especially true in American
19:19Manufacturing Systems where at least
19:22previously to Toyota coming onto the
19:24scene the idea was to produce as much as
19:26possible as fast as possible so what if
19:29the cord is not pulled but something is
19:32wrong in this case the worker committed
19:35a type two error so they failed to
19:39reject the null hypothesis they failed
19:43to reject their assumption that
19:45everything is wrong when something was
19:47actually
19:48wrong so type two error is wrongful
19:51support of the null hypothesis when it
19:54actually should be rejected
19:58now note in this case in this case a
20:02type two error can be
20:05catastrophic it may mean that hundreds
20:08or thousands of cars have that defect
20:12built into them and see this is why in
20:16the Toyota production system and other
20:17Lean Manufacturing Systems pulling that
20:20andon cord is encouraged if not actually
20:25required because the type one error it
20:28an annoyance is okay if they stop the
20:32line for a couple of minutes it's not
20:34that big a deal it's going to mean a few
20:36cars off the end of the line that didn't
20:38produce but think about the reverse if a
20:42worker sees something that they don't
20:44think is right but don't pull the cord
20:47and let that defect keep going and going
20:50and going and then all these cars at the
20:54end of the line have that defect built
20:56in that's why typ one error in this case
21:00is really okay but type two error is
21:04very very bad very bad and if it's a
21:08part on the car that's a safety issue it
21:10can be very very bad so again think of
21:14the difference between type one and type
21:17two error in these two cases type one
21:20was kind of a false alarm so the alarm
21:22was pulled when there really wasn't a
21:24fire the andon cord was pulled when
21:27there really wasn't a major problem
21:29that's annoying but it's not
21:31catastrophic in terms of the
21:32consequences but type two is in the Fire
21:36case if we don't pull the fire alarm and
21:38there is a fire our building may burn
21:42down and people may lose their lives in
21:44this case if we don't pull the andon
21:47cord and there is a problem that defect
21:50may get repeated hundreds or thousands
21:52of times and that's the real world
21:54consequence of type 2 error as related
21:57to type one
22:01error now we can set this up as a couple
22:03of hypotheses like we did the last time
22:05so our null hypothesis is that the
22:07problem is annoying but not serious
22:10everything is okay no major defect so
22:12the normal running of the production
22:14line our null hypothesis is that the
22:17problem is serious everything is not
22:19okay as usual there is a major defect so
22:22we can set up a table just like this so
22:24we have two conclusions we can make we
22:27can accept our n hypothesis or you can
22:29think of it as not rejecting I should
22:31have changed that to not rejecting our
22:33null hypothesis and that is no serious
22:36defect we could reach the conclusion of
22:39rejecting our null hypothesis that there
22:42is some serious defect going on so
22:44that's the conclusion we can make on the
22:46line in the moment now the actual
22:49condition maybe when the supervisor
22:51walks over could be that there is no
22:53serious defect or there is some serious
22:56defect now if we if we do not reject our
23:01null and there is no serious defect
23:04that's the correct conclusion and
23:07actually that's the way the production
23:08line works you know the vast majority of
23:10the time we assume everything is going
23:13okay and there are no serious defects
23:16that's a conclusion that is correct
23:18that's a match now if we reject the no
23:21hypothesis we think there is a serious
23:23defect and the reality is there is a
23:25serious defect that's also correct so
23:28our correct conclusion now let's look at
23:30the type one situation very quickly so
23:32let's say we reject the no hypothesis we
23:35reject our assumption that production is
23:37going okay without serious defect we're
23:40going to reject that because we think
23:41there's a problem now the actual
23:43condition is that there is no serious
23:46problem or defect so that disconnect
23:48between conclusion and reality is the
23:51type one error so the production line
23:53might stop for a few minutes or even
23:55maybe less than a minute to check it
23:58then after we see there's no serious
23:59problem or defect it just starts up
24:02again the type two error that's when we
24:05accept or fail to reject the null
24:08hypothesis so we think everything is
24:10going smoothly when in fact there is a
24:13serious defect going on and that is type
24:15two error and again that gets replicated
24:19over and over and over into each
24:21successive car so in general the real
24:23real consequences of a type two error
24:25are much greater in this case type 2
24:28error would mean hundreds or thousands
24:30of defective cars and maybe even unsafe
24:33cars as we proceed on Down the Line
24:37[Music]

About this transcript

This page contains the full transcript of YouTube transcript (VFMcGdWp0MQ) , generated from the public captions YouTube serves with the video. The transcript has 3,657 words across 531 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.