0:02 all right so we're gonna close out our
0:03 chapter this is the
0:06 last lesson of chapter four and we're
0:08 closing it out by talking about just a
0:10 couple topics that are really
0:12 important in terms of realizing what you
0:13 can do
0:15 and what you can't do with the results
0:17 you get from a study
0:19 so this first slide right here talks
0:21 about a huge
0:24 topic really really important topic that
0:24 you need to
0:28 understand very well called scope
0:32 of inference scope of inference is one
0:33 of those things that
0:35 after you leave my class after you go on
0:36 to be an adult
0:39 hopefully this is a topic that you will remember
0:40 remember
0:42 because it really really matters in
0:44 terms of what you can do
0:46 with results that you get from a study
0:47 and it's something that the average
0:50 adult doesn't understand well enough
0:51 and it's something that if you're
0:53 reading articles online or different things
0:53 things
0:56 about just studies that you see people
0:57 who write articles
1:00 frequently mess this topic up and misapply
1:01 misapply
1:03 the scope of inference and study so i
1:04 have a little bit of a table here at the
1:06 bottom i'm going to go ahead and delete
1:06 that though
1:08 i can't delete it right now but instead
1:09 of filling out this table what i'm going
1:11 to do is have you guys write down
1:13 just the key points of scope of
1:15 inference off on the side somewhere in
1:16 your notes
1:18 there are two things that we are
1:19 concerned about
1:22 with regards to scope of inference we
1:25 care about whether or not our data came
1:31 and we care about whether or not we had random
1:32 random
1:35 assignments these two
1:38 factors are what you need to consider
1:40 in order to decide what you can do with
1:42 your results
1:45 random sampling from a population means
1:47 that you pulled your sample randomly
1:49 out of the entire population everybody
1:51 within that population
1:53 had a chance to be chosen for your study
1:55 if you randomly sample
1:57 you can apply your results to that
1:58 entire population
2:01 if i take a random sample of adults in
2:02 st louis
2:04 i can apply the results from those
2:06 adults to the entire
2:08 population of saint louis couldn't apply
2:09 it to the rest of the u.s
2:11 because i didn't sample from the rest of
2:12 the u.s those people didn't have a
2:14 chance of being chosen
2:18 so random sampling allows you to make
2:21 inference about
2:25 the population inference means basically
2:27 making predictions and stuff like that
2:30 so random sampling allows you to talk
2:32 about the population
2:36 all right random assignment on the other hand
2:36 hand
2:38 is when you break people up and you
2:40 create roughly equivalent groups and experiments
2:41 experiments
2:43 so if you have roughly equivalent groups
2:45 due to random assignments
2:47 and you see a difference between those
2:48 two groups
2:50 you can assume that that difference was
2:52 because of the variable that you manipulated
2:53 manipulated
2:56 random assignment allows you to make conclusions
2:57 conclusions
3:00 or inferences about cause
3:08 so if you want to say that one variable
3:10 causes another one
3:11 it has to be in experiments where you
3:13 have random assignments
3:15 if you want to talk about the population everybody
3:16 everybody
3:18 you need to have random sampling most experiments
3:20 experiments
3:22 that you see in here about are not able
3:24 to be applied to the whole population
3:27 why is that well for experiments when
3:29 you're talking about some college
3:31 research lab or something like that
3:32 they don't randomly pick people out of
3:34 the population and be like you you're
3:36 coming to be in my study i selected you randomly
3:37 randomly
3:40 most experiments rely on volunteers or
3:42 people who sign up to be in the studies
3:43 right there
3:46 so those people that are volunteers are
3:48 probably not representative
3:50 of the population as a whole first off a lot
3:51 lot
3:53 of people who are like subjects in
3:55 experiments are college students
3:56 they're a very common audience right
3:58 there because they're in a university
3:59 anyway they can get extra credit in
4:01 their classes maybe get a little extra money
4:02 money
4:03 so a lot of experiments who are
4:05 performed on college kids
4:07 they're typically younger healthier than
4:09 the average person in our population
4:12 so we can't apply the results that we
4:13 see in college kids
4:15 to the rest of the population if we're
4:17 taking some group that's like high
4:19 cholesterol risk and we're doing an
4:20 experiment on them
4:22 we can only talk about how this drug
4:24 works in people with high cholesterol
4:26 like the ones in the study so the
4:28 disclaimer you usually see in experiments
4:28 experiments
4:31 is that the results can be applied we
4:32 can determine that this medication
4:35 caused the benefits or the whatever but
4:38 we can only apply it to people like
4:40 those in the study so it's like a little
4:42 disclaimer in most experiments that you
4:43 will see
4:44 so what this table is doing is just
4:46 summarizing what i've been talking about
4:47 right here
4:49 were they randomly selected were they
4:51 randomly assigned
4:54 if you are randomly selected that means
4:54 that you
4:58 can make inferences about the population
5:01 okay so were they randomly selected yes
5:03 yes you can talk about the population
5:05 yes you can talk about the population
5:07 if they are not randomly selected that
5:10 means that you cannot talk
5:14 about the population were they randomly assigned
5:15 assigned
5:18 treatments if yes then you can
5:22 talk about cause and effects and if no
5:25 you cannot talk about cause and effects
5:26 so you can see it parenthetically right here
5:27 here
5:29 most experiments fall into this group
5:30 you can't make inferences about the
5:31 population because they're mostly
5:33 volunteers for the study
5:35 but you can talk about cause and effects
5:37 a lot of observational studies fall over
5:39 here where there was a random sample you
5:40 can talk about the population
5:43 but you can't talk about cause and fact
5:45 flashback to one of our earlier lessons
5:47 where we looked at smoking compared to
5:49 adhd rates in children
5:51 that wasn't an experiment so what we
5:52 could say
5:54 in that context is that it appears that
5:56 smoking is connected
5:58 with higher rates of adhd those two
5:59 things go together
6:02 and if you smoke you're probably also
6:03 going to have a higher risk of having a
6:05 kid with adhd
6:08 but we can't go that extra step to say
6:09 the smoking caused
6:12 the rate of adhd to increase because of
6:14 confounding factors we don't know what's
6:16 actually responsible for it
6:18 newspaper articles mess up scope of
6:20 inference all the time
6:22 you will see studies all the time where
6:24 maybe the researcher knew what they were
6:25 doing and the researcher didn't do
6:26 anything wrong
6:28 but some journalist who reads it or some
6:29 person who wants to make their buzzfeed
6:31 article about whatever
6:33 doesn't understand scope of inference so
6:35 they incorrectly interpret the results
6:36 to say oh this
6:39 caused this we'll look at some articles
6:42 where that happens in class
6:44 so what i have here is just an example
6:47 to practice scope of inference here
6:48 we're going to test whether listening to
6:50 music while you
6:53 work um impacts your i think it's going
6:54 to be grades
6:56 in school gpa at the end of the semester
6:59 okay and what i have right here
7:01 is four hypothetical designs and we're
7:03 just going to talk really quickly these
7:04 are pretty easy questions
7:06 once you understand scope of inference
7:08 where can we apply our results
7:10 okay so the following slide right here
7:12 is where you can write these down but
7:13 i'm just going to do it on the slide so
7:15 i don't have to tab back and forth i'm
7:16 going to write it right here
7:19 okay so we are going to take everybody
7:20 in our ap stats class
7:22 and we are going to have them be in a
7:25 study ask them hey do you guys listen to
7:28 music or not and then look at gpa
7:30 um at the end of the semester so i let
7:32 you guys decide do you listen to music
7:34 people who say yeah are in this group
7:35 people who say no are in this group end
7:38 of semester we would look at them
7:42 okay so we did not randomly sample
7:45 we used an ap stats class we can't talk
7:47 about the whole population
7:52 so we can't talk about
7:59 only this class so if there is a
8:00 difference between
8:04 the two groups i can say in this class
8:06 kids who listen to music do better or do
8:07 worse that would be fair
8:09 but i can't apply my results beyond my
8:10 classroom because i
8:14 didn't randomly sample okay i also
8:15 didn't randomly assign you
8:18 to groups since you were not randomly
8:19 assigned to groups
8:21 that means if there's a difference i
8:22 can't say it was because of the music
8:24 because there could be other confounding
8:26 factors like um
8:29 maybe kids who like music are
8:32 more motivated or less motivated you can
8:34 run all those hypotheticals
8:36 so we can't talk about cause and effect
8:40 either so we can't say
8:43 the music caused
8:46 a difference the word caused
8:50 is a very powerful word in ap stats
8:52 don't use it unless you actually have
8:54 cause and effect you can talk about
8:54 things being
8:56 linked together or there being a
8:58 connection but mentioning
9:01 effect or cause is a really strong
9:02 statement that you should not do unless
9:04 you were randomly sampling
9:06 okay so this is how i would want you
9:07 guys to write your answers but i'm going
9:09 to write them a lot shorter
9:10 in these next couple examples to not
9:12 slow things down right here
9:14 now in the second scenario we are
9:17 randomly sampling from our school
9:18 and then we ask if they listen to music
9:20 or not and be breaking unity groups
9:23 so i did randomly assemble that means i can
9:25 can
9:29 apply to the population
9:32 of my school can't apply beyond my
9:33 school because i didn't use people from
9:34 other schools
9:36 but i can apply to the whole school even
9:38 if i don't talk to every single person
9:42 but there would be no cause and effects
9:43 so i could say the same sort of thing
9:45 can't say that music caused a difference
9:46 because i just let you establish your
9:49 own group same as the first one
9:52 all right next up get every nap stats
9:54 class so you guys are my captive
9:55 audience you're in my study
9:57 now i'm going to randomly assign half of
9:59 you to each group
10:01 random assignments should make it so the
10:03 groups are about equal
10:04 in all those different factors you can
10:06 think about so if there's a difference
10:09 i can say it's due to the music so this
10:19 but a yes to cause and effect
10:20 and then finally on this last one right
10:22 here this is like the gold standard it's
10:24 a random sample from my school
10:26 and i did do random assignments so i
10:29 could make conclusions about both
10:31 this is generally unrealistic in most experiments
10:32 experiments
10:34 like i said experiments tend to live
10:36 here observational studies
10:41 so the last thing i believe i need to
10:42 talk about with you guys
10:45 oh two things okay first of all this is
10:46 an article we're probably going to read
10:47 in class but it's kind of interesting
10:50 um based on taking vitamins and
10:53 the connection that vitamins have with
10:55 your overall health
10:58 turns out that when you do experiments
10:59 where you actually randomly assign
11:01 people to take vitamins or knots
11:03 there's usually not too much of a
11:05 difference in health outcomes
11:07 between the two groups vitamins don't
11:08 tend to do a lot
11:12 in terms of your overall health but
11:13 if you do an observational study where
11:15 you don't do random assignment you just
11:16 look at data
11:19 people who take vitamins tend to have
11:21 better health outcomes than people who don't
11:21 don't
11:24 but it's because of confounding factors
11:26 if you're taking vitamins you probably
11:27 care more about your health you do other
11:30 things like exercise diet et cetera
11:32 that make you have those better health
11:33 outcomes so if you're not
11:36 careful you can very easily run into an issue
11:36 issue
11:38 where those confounding factors make you
11:40 think something is causing
11:44 a situation when it's actually not
11:46 and the very last thing that we have to
11:47 talk about in this chapter
11:50 it's quick but it's important being ethical
11:51 ethical
11:53 and collecting your data just something
11:55 that we have to talk about here
11:56 um because you guys will be running your
11:59 own project later on this semester
12:01 whenever you collect data you have an obligation
12:02 obligation
12:03 not to share that data with people
12:05 outside your studies and keep things
12:08 confidential you have to be upfront with
12:10 people you can't lie to your um the
12:11 people in your study
12:13 and you can't be like oh this person
12:14 said this on their survey
12:17 and that's just not an okay thing to do
12:18 um when you if you actually go into a
12:20 career with research and you violate
12:22 these principles of ethics you can
12:23 actually be like
12:26 blacklisted from the um like community
12:28 so you can't actually participate in
12:29 research further
12:32 um all studies if you guys do any like
12:32 high level
12:35 statistical studies you'll submit them
12:37 to a review board and they'll make very
12:39 certain that you're not breaking any
12:41 guidelines there so use common sense and
12:42 common decency
12:44 when you guys do your project for me
12:45 later this semester
12:48 results do need to be kept confidential
12:49 so that
12:51 is the end of our first unit on experimental
12:53 experimental design
