WEBVTT Kind: captions; language: en-us NOTE Treffsikkerhet: 81% (HØY) 00:00:02.400 --> 00:00:10.500 This is the second part of a presentation to prepare you for this SPED4010 course. And in this 00:00:10.500 --> 00:00:18.300 part, I am going to say a few general things about what statistics is. NOTE Treffsikkerhet: 91% (HØY) 00:00:18.300 --> 00:00:31.400 So in general we can say that statistics is a tool in or a set of procedures and the role of this 00:00:31.400 --> 00:00:40.800 tool is to help us manage and handle uncertainty and variability. So is the way that we have to deal 00:00:40.800 --> 00:00:49.150 with the fact that things are uncertain and things and data are variable, people are variable. NOTE Treffsikkerhet: 91% (HØY) 00:00:49.150 --> 00:00:56.900 And responses are variable and the numbers we obtain our variable and the they are uncertain and the 00:00:56.900 --> 00:01:04.800 performance of anyone and anything is uncertain. And we just need to be able to handle this idea 00:01:04.800 --> 00:01:12.950 that there is not perfect consistency. And where is this uncertainty? Well everywhere, so everything 00:01:12.950 --> 00:01:17.949 in reality in real life is uncertain and variable. NOTE Treffsikkerhet: 86% (HØY) 00:01:17.949 --> 00:01:26.000 The only place where there is no uncertainty or variability is in math in. Well, not every branch of 00:01:26.000 --> 00:01:34.300 math because statistics is also comes from a branch of mathematics, but you can find mathematical 00:01:34.300 --> 00:01:41.450 domains where things are certain and logic is another example, but everywhere else in every real 00:01:41.450 --> 00:01:47.750 part of life. Everything is uncertain and variable. The only tool we have to deal with it in NOTE Treffsikkerhet: 91% (HØY) 00:01:47.750 --> 00:01:56.250 way that leads to justified decisions is statistics. So statistics NOTE Treffsikkerhet: 90% (HØY) 00:01:56.250 --> 00:02:06.100 gives us the possibility to evaluate all sorts of situations in which, of course, we are faced with 00:02:06.100 --> 00:02:14.000 uncertainty and variability, so that we can assess how much confidence can be put in proposals for 00:02:14.000 --> 00:02:20.700 doing things differently. In results that are reported in findings that are reported in guidelines, 00:02:20.700 --> 00:02:25.550 that are proposed methods decisions about our lives and so on. NOTE Treffsikkerhet: 89% (HØY) 00:02:25.550 --> 00:02:32.550 Statistics is indeed based on a mathematical branch of probability theory. NOTE Treffsikkerhet: 87% (HØY) 00:02:32.550 --> 00:02:40.400 As I said in the previous part of this presentation, we are not going to deal with this math. So the 00:02:40.400 --> 00:02:46.400 mathematicians and statisticians, develop the math that is behind the methods that were using in or 00:02:46.400 --> 00:02:52.700 just using them. What do we need is only to understand how to use them and how to interpret them. So 00:02:52.700 --> 00:02:56.000 we have to be focusing on the concepts. NOTE Treffsikkerhet: 79% (HØY) 00:02:56.400 --> 00:03:04.600 Let us go back to the examples of part 1 to say just a few more words about what statistics does in 00:03:04.600 --> 00:03:09.250 helping us deal with each of these hypothetical situations. NOTE Treffsikkerhet: 91% (HØY) 00:03:09.250 --> 00:03:16.500 So, in the first example, there was talk of specific test scores, and there was the example of a 00:03:16.500 --> 00:03:21.250 score, a standard score of 69, in receptive vocabulary. NOTE Treffsikkerhet: 91% (HØY) 00:03:21.250 --> 00:03:30.300 So what has standard score mean and what is 69? In order to be able to interpret this report you 00:03:30.300 --> 00:03:39.500 need to know about uncertainty in the score. So what if we tested this same child with the same test 00:03:39.500 --> 00:03:46.700 again, would you get the same score will almost certainly not. How far can we expect the second score 00:03:46.700 --> 00:03:51.100 to be from the first one? That is an inherent uncertainty in the test. NOTE Treffsikkerhet: 89% (HØY) 00:03:51.100 --> 00:03:57.100 Then what if we tested with the different vocabulary score with a different vocabulary test? Would 00:03:57.100 --> 00:04:04.149 it get the same score? Almost certainly not. Then we need to know how much variability we have. 00:04:04.149 --> 00:04:07.649 that is a function of which test we chose to use. NOTE Treffsikkerhet: 91% (HØY) 00:04:07.649 --> 00:04:15.399 And then, in order to understand what 69 means, we need to know something about, not the test so 00:04:15.399 --> 00:04:23.500 much, but the population. So if we were to give this test to every child, how many children would 00:04:23.500 --> 00:04:30.100 get more than sixty nine. How many would get less than sixty nine. This is the basis on which you 00:04:30.100 --> 00:04:37.300 can say whether this particular kit that we are dealing with has a low score or a very low score or 00:04:37.300 --> 00:04:37.900 an okay score. NOTE Treffsikkerhet: 84% (HØY) 00:04:37.900 --> 00:04:43.950 and what that means. So this is all about uncertainties and variabilities. NOTE Treffsikkerhet: 89% (HØY) 00:04:43.950 --> 00:04:47.700 Moving on to the second example. NOTE Treffsikkerhet: 91% (HØY) 00:04:48.000 --> 00:04:56.300 With this screening outcome. So, there was this screening procedure that is supposed to identify 00:04:56.300 --> 00:05:01.200 children with dyslexia. Well, first of all, we need to be able to know NOTE Treffsikkerhet: 88% (HØY) 00:05:01.200 --> 00:05:09.400 something about the uncertainty of the outcome itself. So, if we were giving this screening test to 00:05:09.400 --> 00:05:17.300 children with dyslexia and the children without dyslexia, how often would the test get it, right? 00:05:17.300 --> 00:05:23.300 So, if we were only giving the test to children with dyslexia, would they all be identified as 00:05:23.300 --> 00:05:29.700 having as an in need of follow-up? If we're giving these tests of children, without dyslexia, 00:05:29.700 --> 00:05:31.750 would the test always say this NOTE Treffsikkerhet: 91% (HØY) 00:05:31.750 --> 00:05:37.400 kids are okay. They don't need to follow up for dyslexia. So false, positives and false negatives 00:05:37.400 --> 00:05:43.600 are very important. That has to do with how well the test is performing, but that's not the whole 00:05:43.600 --> 00:05:53.200 story. Again. We need to know something about variability in the population. So what if we are 00:05:53.200 --> 00:06:01.200 screening for something that we expect about 5% of the population to be identified like dyslexia. NOTE Treffsikkerhet: 80% (HØY) 00:06:01.200 --> 00:06:08.000 Or maybe we should be, we might want to identify maybe 10% if we want to be able to help children, 00:06:08.000 --> 00:06:12.200 who are kind of in the in the gray zone. NOTE Treffsikkerhet: 91% (HØY) 00:06:12.200 --> 00:06:19.200 There are other kinds of tests where we need to identify people who are much rarer. So, in medical 00:06:19.200 --> 00:06:25.200 conditions, we may want to have screening test for something that occurs in every, in one person, 00:06:25.200 --> 00:06:32.700 every thousand or every 10,000. So very rare conditions. How do screening tests perform in such 00:06:32.700 --> 00:06:33.850 cases. NOTE Treffsikkerhet: 91% (HØY) 00:06:33.850 --> 00:06:41.700 This is an issue about the variability of the population that combines with the uncertainty issues 00:06:41.700 --> 00:06:48.900 about the test itself and only one we understand both can we interpret results from this screening 00:06:48.900 --> 00:06:54.800 test and be able to decide whether we should use it or not, taking into account all the pros and cons 00:06:54.800 --> 00:07:02.500 of it. So given the special circumstances of our situation today. I thought to recommend you to 00:07:02.500 --> 00:07:04.000 listen to NOTE Treffsikkerhet: 78% (HØY) 00:07:04.000 --> 00:07:11.600 read and listen to some information about testing that are general. Actually, these are about 00:07:11.600 --> 00:07:19.800 medical testing. But whenever we have screening and diagnosis, whether it's medical as in the 00:07:19.800 --> 00:07:25.500 examples that I've linked to here or educational as in dyslexia, which is not a medical issue. In 00:07:25.500 --> 00:07:31.000 any sense at all, but are using similar concepts. We using responsive, like, screening. We are using 00:07:31.000 --> 00:07:34.000 the concept of diagnosis. So these kinds of NOTE Treffsikkerhet: 84% (HØY) 00:07:34.000 --> 00:07:39.400 concerns come up and you can check out these links sources. The videos, these videos are very short. 00:07:39.400 --> 00:07:44.700 There are few minutes each, so you can see the kinds of concerns that I'm talking about and how they 00:07:44.700 --> 00:07:49.900 must be understood in order to be able to interpret findings of screening tests. NOTE Treffsikkerhet: 81% (HØY) 00:07:50.200 --> 00:07:59.850 I our third example, we had an announced Intervention, which was said to be highly effective. NOTE Treffsikkerhet: 91% (HØY) 00:07:59.850 --> 00:08:10.150 Now, this is a very contentious issue. And in order to decide whether this is worth adopting, you 00:08:10.150 --> 00:08:16.500 need to understand what a research study like the one in this hypothetical example, what such a 00:08:16.500 --> 00:08:22.200 research study shows and this is counterintuitive. NOTE Treffsikkerhet: 88% (HØY) 00:08:22.200 --> 00:08:29.900 Because what you first need to understand is what happens when you're sampling a group of 00:08:29.900 --> 00:08:38.950 individuals from a population. So here remember this was an announced intervention for difficulties 00:08:38.950 --> 00:08:46.700 in math. So there was a study of kids who received this intervention and improve their math skills 00:08:46.700 --> 00:08:48.700 as measured by a test. NOTE Treffsikkerhet: 84% (HØY) 00:08:48.700 --> 00:08:56.750 So the hypothetical researchers who did this, they pick 10 kids to run the intervention on. NOTE Treffsikkerhet: 84% (HØY) 00:08:56.750 --> 00:09:06.250 And this only makes sense if we assume that, what they find with these 10 kids can be valid to 00:09:06.250 --> 00:09:09.500 generalize over the whole population of children. NOTE Treffsikkerhet: 77% (HØY) 00:09:09.500 --> 00:09:16.300 Now, I didn't tell you in the example, if they chose 10 children with difficulty or with a little 00:09:16.300 --> 00:09:21.600 difficulty, with a lot of difficulty or just ten random children, but that's the kind of information 00:09:21.600 --> 00:09:29.500 about variability that I didn't give you and that is critical to interpreting the test, but there is 00:09:29.500 --> 00:09:32.250 also the more general effect. So if you NOTE Treffsikkerhet: 82% (HØY) 00:09:32.250 --> 00:09:39.800 have a large population of hundreds of thousands of children, and you pull out 10 of them. NOTE Treffsikkerhet: 85% (HØY) 00:09:40.300 --> 00:09:46.600 And then you pull out a different set of 10. Are you going to get the same answer from your two 00:09:46.600 --> 00:09:53.000 samples? Probably not how different in the answer is be. This is very important because that helps 00:09:53.000 --> 00:09:59.500 you evaluate how much trust you can put into the answer from any single sample. Every study has one 00:09:59.500 --> 00:10:08.900 sample, but the weight interpreted properly is to think what could I expect if they had another 00:10:08.900 --> 00:10:10.050 sample? NOTE Treffsikkerhet: 91% (HØY) 00:10:10.050 --> 00:10:17.000 So understanding variability, among different individuals children, and therefore among different 00:10:17.000 --> 00:10:25.400 samples from the same population is critical in being able to guide us to interpret finding such as 00:10:25.400 --> 00:10:32.700 these of intervention. Can we assume that a different sample of children would also improve in their 00:10:32.700 --> 00:10:36.350 math skills? And then how much would each child improve? NOTE Treffsikkerhet: 76% (HØY) 00:10:36.350 --> 00:10:44.400 To say that they improved on average by 15% could mean that some children improved by 30% and some 00:10:44.400 --> 00:10:50.600 children didn't improve or that some children really improved very much and some children even got 00:10:50.600 --> 00:10:59.300 worse. So you can't know by just having an average estimate of improvement in a single test. Those 00:10:59.300 --> 00:11:04.800 are all these issues of uncertainty in variability that go in, that you really need to NOTE Treffsikkerhet: 86% (HØY) 00:11:04.800 --> 00:11:13.600 have in the assessment of whether this kind of intervention is worth adopting or not. NOTE Treffsikkerhet: 90% (HØY) 00:11:13.900 --> 00:11:22.400 Moving on to the fourth example. That was the one where you had an idea about an intervention and 00:11:22.400 --> 00:11:24.600 you tried it on 20 kids. NOTE Treffsikkerhet: 77% (HØY) 00:11:24.600 --> 00:11:28.400 And 13 of them proved and seven didn't. NOTE Treffsikkerhet: 91% (HØY) 00:11:28.400 --> 00:11:32.500 So the question is was this a good idea? NOTE Treffsikkerhet: 91% (HØY) 00:11:32.500 --> 00:11:40.800 Now again, this is a little counterintuitive but to answer this question. The first thing you would 00:11:40.800 --> 00:11:47.200 do from a statistical point of view is not to go and study your intervention. But is to try to 00:11:47.200 --> 00:11:55.400 understand whether this kind of difference should be thought of as unexpected. NOTE Treffsikkerhet: 86% (HØY) 00:11:55.400 --> 00:12:05.000 So is it unexpected to get 13 out of 20 to improve? Is it unexpected by chance? What does it mean by 00:12:05.000 --> 00:12:09.600 chance? By chance, means if nothing is really going on. NOTE Treffsikkerhet: 91% (HØY) 00:12:09.600 --> 00:12:19.100 So if I have 20 kids, I measure them on some skill that's of interest to me for my work and then I 00:12:19.100 --> 00:12:23.450 do nothing and then I measure them again on the same skill. NOTE Treffsikkerhet: 91% (HØY) 00:12:23.450 --> 00:12:30.200 So there will be some random variability that will depend on the reliability of the test. So the kid 00:12:30.200 --> 00:12:40.000 will not get the same score on both assessment on both assessment time points. And therefore, some 00:12:40.000 --> 00:12:44.100 of them may get a higher score in second time. Some of them may get a lower score. This is 00:12:44.100 --> 00:12:47.700 completely normal and expected and you shouldn't be surprised by it. NOTE Treffsikkerhet: 86% (HØY) 00:12:47.700 --> 00:12:55.650 So if you got 13 to have a higher score, the second time in seven have a lower score. NOTE Treffsikkerhet: 91% (HØY) 00:12:55.650 --> 00:12:58.350 Would that be weird? NOTE Treffsikkerhet: 91% (HØY) 00:12:58.350 --> 00:13:06.900 If nothing happened, so if that would be weird, if that would be unexpected. Then it means. Well, if 00:13:06.900 --> 00:13:11.200 intervention got me to that point, then maybe it's a good one. NOTE Treffsikkerhet: 83% (HØY) 00:13:11.200 --> 00:13:19.900 But if getting this difference by complete chance, by just reading the test is not unexpected. Then 00:13:19.900 --> 00:13:25.700 well, you can't say very much about intervention. It doesn't sound so promising after all. So the 00:13:25.700 --> 00:13:28.050 critical idea here is NOTE Treffsikkerhet: 87% (HØY) 00:13:28.050 --> 00:13:37.000 chance. Imagine you flip a coin, 20 times, and you get heads 13 out of these 20 times not in a row, 00:13:37.000 --> 00:13:44.900 but just any 13 of them. And, of course, the other seven are tails, is that impressive is that 00:13:44.900 --> 00:13:51.700 unexpected without completely normal and, you know, not remarkable in any way? NOTE Treffsikkerhet: 84% (HØY) 00:13:51.700 --> 00:14:04.600 So if you don't find 13 heads in 20 flips impressive, why would you find 13 improvements in 20 kids 00:14:04.600 --> 00:14:13.800 impressive? That's the way statistics goes, kind of in around, in a counter intuitive way of 00:14:13.800 --> 00:14:20.500 thinking, to highlight the role of chance and why is there all of chance important? Because 00:14:20.500 --> 00:14:22.200 everything is uncertain and NOTE Treffsikkerhet: 73% (MEDIUM) 00:14:22.200 --> 00:14:23.600 everything is variable. NOTE Treffsikkerhet: 91% (HØY) 00:14:23.600 --> 00:14:30.200 And that we know for certain that everything is uncertain. Okay, that sounded a little strange. It 00:14:30.200 --> 00:14:38.000 actually is true. So if we know that everything is uncertain, then we have to learn and tune our 00:14:38.000 --> 00:14:48.800 intuition to appreciate what counts as unexpected, surprising or remarkable. And that's where we see 00:14:48.800 --> 00:14:53.849 the value of things. Possibly having made a difference. Because if NOTE Treffsikkerhet: 78% (HØY) 00:14:53.849 --> 00:14:59.100 something is unremarkable, then we have no reason to think that something made a difference. So if 00:14:59.100 --> 00:15:05.900 you got the coin to come up heads 13 out of 20 times, you wouldn't think someone is tricking you you 00:15:05.900 --> 00:15:13.000 wouldn't think this is a strange coin. A fair coin with a 50/50 chance of heads and tails will 00:15:13.000 --> 00:15:20.300 pretty normally come up 13 or so or 12 or 10 or 14 times out of 20 heads, and you wouldn't be 00:15:20.300 --> 00:15:23.900 worried. If you got 19 out of 20 head NOTE Treffsikkerhet: 79% (HØY) 00:15:24.700 --> 00:15:31.250 then you might be worried like what's wrong with this coin. This is coin is strange. But 13. Well 00:15:31.250 --> 00:15:38.100 not so worrisome. Well, then maybe you shouldn't be so impressed by this fictional intervention in 00:15:38.100 --> 00:15:48.100 this example. So going back to the more general description of what statistics does based on some 00:15:48.100 --> 00:15:54.000 intuitions hopefully drawn out of these examples, what statistics does for you is to give you 00:15:54.000 --> 00:15:54.400 procedures NOTE Treffsikkerhet: 85% (HØY) 00:15:54.400 --> 00:16:02.600 by which you can calculate results. Of course, it also gives you a way of thinking the, 00:16:02.600 --> 00:16:08.500 what kinds of things you should be calculating to get your answer. And although the procedures are 00:16:08.500 --> 00:16:15.600 straightforward. The thinking behind the procedures is not entirely intuitive and that's why some 00:16:15.600 --> 00:16:21.450 people find statistics hard. It's not that the math is hard. So people tend to be scared by numbers, 00:16:21.450 --> 00:16:24.400 but the numbers aren't scary at all in statistics. NOTE Treffsikkerhet: 90% (HØY) 00:16:24.400 --> 00:16:31.000 Nothing is really scary. But the way of thinking is unfamiliar. Do you really need to change your 00:16:31.000 --> 00:16:38.800 way of thinking and go in this roundabout way to understand what should be calculated and why? Now 00:16:38.800 --> 00:16:47.500 statistics doesn't just give you results. It always produces two kinds of things. Two parts of an 00:16:47.500 --> 00:16:54.150 answer together. That's part of doing a statistical calculation. You get an estimate. NOTE Treffsikkerhet: 82% (HØY) 00:16:54.150 --> 00:17:00.600 And you get a confidence that's associated with this estimate. So, for every answer you get out of 00:17:00.600 --> 00:17:06.800 Statistics, you also get the degree of certainty associated with it. How much can you trust the 00:17:06.800 --> 00:17:13.300 answer? And that's the real strength, an asset of Statistics. That's why use it. We don't use it to 00:17:13.300 --> 00:17:16.050 find the number that's our answer. NOTE Treffsikkerhet: 76% (HØY) 00:17:16.050 --> 00:17:23.700 That's not really very informative because that doesn't tell us anything about the uncertainty 00:17:23.700 --> 00:17:32.000 associated with the answer. So if the answer is two, but could be anything between one and three, 00:17:32.000 --> 00:17:38.250 that's very different from a case in which the answer is 2, but could be anything between 0 and 20. 00:17:38.250 --> 00:17:44.800 So the first answer is much, more useful. We know it's about 2. The second case, if it's 2, 00:17:44.800 --> 00:17:45.950 but could have been zero or NOTE Treffsikkerhet: 85% (HØY) 00:17:45.950 --> 00:17:51.600 Could have been 20. That's not really useful answer. Just, that's what statistics does for us. Gives 00:17:51.600 --> 00:17:59.000 us a number and the confidence associated with that number. And the confidence is actually given in 00:17:59.000 --> 00:18:04.800 the form of a range. And you're already familiar with the idea of a range because you see that, for 00:18:04.800 --> 00:18:12.500 example, in polls in elections, that they say, the so, so party is predicted to get so-and-so 00:18:12.500 --> 00:18:16.000 percentage of the vote. And they say, the margin is NOTE Treffsikkerhet: 90% (HØY) 00:18:16.000 --> 00:18:24.900 - 1% of plus or minus 3%. So you're already familiar with the idea that a statistical estimate comes 00:18:24.900 --> 00:18:30.600 with a range. So when they say that a party will get 20% of the vote. And there's the footnote that 00:18:30.600 --> 00:18:38.600 this, there's a 2% range. You know, okay, it's 20%. But it could be 18, could be 22 and the 00:18:38.600 --> 00:18:42.500 prediction would be correct anywhere in that interval. NOTE Treffsikkerhet: 85% (HØY) 00:18:43.500 --> 00:18:54.000 So, that would be different from an another poll that says, predict 20, with an error range of plus 00:18:54.000 --> 00:19:02.600 or minus 5, so they would predict between 15 and 25 verses predicting between 18 and 22. So the two 00:19:02.600 --> 00:19:11.700 poles are clearly of different value, which means statistics gives you a sense of a likely answer. NOTE Treffsikkerhet: 74% (MEDIUM) 00:19:11.700 --> 00:19:15.250 Plus a sense of how much you can trust that. NOTE Treffsikkerhet: 83% (HØY) 00:19:15.250 --> 00:19:22.300 And if this is usually expressed as intervals, so a interval is a range of values that are 00:19:22.300 --> 00:19:28.900 likely. And this is also a something that you need to start thinking about this new way of 00:19:28.900 --> 00:19:32.700 approaching answers, were used to answers as being points. NOTE Treffsikkerhet: 84% (HØY) 00:19:32.700 --> 00:19:39.300 Point is a specific number. So indeed because people are more used to points and points are easier 00:19:39.300 --> 00:19:44.700 to process the the polls tells us. Okay, the party's going to get twenty percent. So that's a point. 00:19:44.700 --> 00:19:51.050 That's a specific number, but that's not what this study actually showed. The study showed between 00:19:51.050 --> 00:19:57.300 18 and 22. They give you the 20 as a point and they also give you the confidence interval. Say, 00:19:57.300 --> 00:20:00.300 okay. They started really is a plus or minus 2. NOTE Treffsikkerhet: 88% (HØY) 00:20:00.300 --> 00:20:05.200 And the statistical way of understanding. This result is to think about the interval, not about the 00:20:05.200 --> 00:20:06.350 point. NOTE Treffsikkerhet: 89% (HØY) 00:20:06.350 --> 00:20:15.800 And even when you have this tool the statistical tool, of course, it's still up to you to be able to 00:20:15.800 --> 00:20:22.500 choose the right procedure. So unfortunately, statistics is very rich. This means there are many 00:20:22.500 --> 00:20:27.900 different ways to do very different things. And you have to know which is right for the kind of 00:20:27.900 --> 00:20:33.100 question that you have for the kind of data set up that you have. And you also need to be able to 00:20:33.100 --> 00:20:36.900 interpret the answer that you get out of Statistics. So it's like a toolbox. NOTE Treffsikkerhet: 90% (HØY) 00:20:36.900 --> 00:20:44.000 If you don't know how to use a hammer and a screwdriver, then you may be hitting with a screw and 00:20:44.000 --> 00:20:48.700 that's not a problem with the hammer or the screwdriver. That's the problem of not knowing how to 00:20:48.700 --> 00:20:53.300 use the tools. So, there is a similar issue with Statistics and that's why it's important to 00:20:53.300 --> 00:20:58.400 understand the procedures, and when they apply, and what they mean. NOTE Treffsikkerhet: 75% (MEDIUM) 00:21:00.100 --> 00:21:09.850 This is related to the more general context of research methods, which is the whole topic of 4010 00:21:09.850 --> 00:21:16.600 and has to do with what counts as a scientific question. So, what kinds of questions can statistics 00:21:16.600 --> 00:21:23.200 help you answer? The first thing that we need to understand is that hard questions can only be 00:21:23.200 --> 00:21:29.400 answered if they are sufficiently well specified. So what is it mean "Sufficiently well specified"? 00:21:29.400 --> 00:21:30.950 We will see an example of that. NOTE Treffsikkerhet: 91% (HØY) 00:21:30.950 --> 00:21:40.200 They shouldn't be vague. They most of the questions that we have are actually not well specified. 00:21:40.200 --> 00:21:47.450 Because the only question is that statistics can help you answer is hypotheses about relationships 00:21:47.450 --> 00:21:53.600 among variables. So that probably had a few unknown words in there. And we're going to spend some 00:21:53.600 --> 00:21:56.150 time in the course to explain those. NOTE Treffsikkerhet: 91% (HØY) 00:21:56.150 --> 00:22:05.800 In particular, the study of relationships among variables. It's how every branch of science works. 00:22:05.800 --> 00:22:12.800 So that's the physical sciences, the life sciences and the social sciences. So everything we can say 00:22:12.800 --> 00:22:15.950 that we know or know something about NOTE Treffsikkerhet: 91% (HØY) 00:22:15.950 --> 00:22:23.500 it's because questions were posed in terms of relationships, relationships among variables 00:22:23.500 --> 00:22:31.150 variables were measured data were analyzed and answers were provided regarding these relationships. 00:22:31.150 --> 00:22:37.550 And these relationships in statistics often take the form of prediction. NOTE Treffsikkerhet: 91% (HØY) 00:22:37.550 --> 00:22:41.850 Predicting one measure from another. NOTE Treffsikkerhet: 91% (HØY) 00:22:41.850 --> 00:22:48.400 This is a little counterintuitive because that's not how we normally think about predict. NOTE Treffsikkerhet: 84% (HØY) 00:22:48.400 --> 00:22:55.500 So we normally think about predictors having to do with the future. Like it pulled predicts voting 00:22:55.500 --> 00:23:03.100 out comes the this only makes sense if the ball was made before the voting day. Otherwise, you just 00:23:03.100 --> 00:23:11.300 count the votes. So in that sense, that's it actual predictive use of Statistics where you ask some 00:23:11.300 --> 00:23:15.400 people, what they're going to vote. And then you predict what the actual election outcome is going 00:23:15.400 --> 00:23:18.050 to be. But in statistics the word prediction. NOTE Treffsikkerhet: 79% (HØY) 00:23:18.050 --> 00:23:24.050 isn't always about the future. We can say we can predict NOTE Treffsikkerhet: 75% (MEDIUM) 00:23:24.050 --> 00:23:28.199 how will you will do at school NOTE Treffsikkerhet: 84% (HØY) 00:23:28.199 --> 00:23:36.500 Based on knowing something about you in kindergarten. So that's again a prediction. Right? That's 00:23:36.500 --> 00:23:42.400 it. It, it usual, the usual sense of the sense of the word prediction. You measure something in 00:23:42.400 --> 00:23:49.000 kindergarten. Try to predict, which kid will do well in school, but you can also go and say, 00:23:49.000 --> 00:23:55.100 second or third grade. It's specific time point. And say, can I predict how well this kid is doing 00:23:55.100 --> 00:23:58.500 in math, by knowing their language, at the same time NOTE Treffsikkerhet: 91% (HØY) 00:23:58.500 --> 00:24:05.200 point, not a previous standpoint, can I predict how well this kid is doing in reading by knowing their 00:24:05.200 --> 00:24:07.200 vocabulary for example. NOTE Treffsikkerhet: 88% (HØY) 00:24:07.200 --> 00:24:17.700 So this is trying to see if you can predict, if you can guess correctly, the values in one 00:24:17.700 --> 00:24:24.650 variable, for example, reading performance by knowing the values on another variable say vocabulary. 00:24:24.650 --> 00:24:31.550 This is important because if the answer is yes, if vocabulary and reading performance are related. NOTE Treffsikkerhet: 82% (HØY) 00:24:31.550 --> 00:24:38.000 Then that's useful information. That's the answer that you might be interested in. And then of 00:24:38.000 --> 00:24:42.900 course, how you interpreted, and how you act upon it is not a statistical issue, but that's a kind 00:24:42.900 --> 00:24:49.600 of prediction and that shows how, what we mean when we say relationships among variables. So even an 00:24:49.600 --> 00:24:52.950 intervention is a relationship among variables. NOTE Treffsikkerhet: 90% (HØY) 00:24:52.950 --> 00:24:56.650 So what does this mean? Can I predict NOTE Treffsikkerhet: 91% (HØY) 00:24:56.650 --> 00:25:05.300 how well a child will do in a math test. If I know whether this child has received an intervention 00:25:05.300 --> 00:25:11.100 or not. So one variable is intervention or not. The other variable is a math test this relates to 00:25:11.100 --> 00:25:12.900 our previous example. NOTE Treffsikkerhet: 81% (HØY) 00:25:12.900 --> 00:25:19.900 If I can predict, if actually, I can guess better NOTE Treffsikkerhet: 89% (HØY) 00:25:19.900 --> 00:25:22.350 a child's math score NOTE Treffsikkerhet: 86% (HØY) 00:25:22.350 --> 00:25:29.800 when I know whether or not they have received intervention compared to, when I don't know whether or 00:25:29.800 --> 00:25:35.300 not, they have received an adventure, if the guest is improved by knowing about the intervention, 00:25:35.300 --> 00:25:39.750 this, this is evidence that intervention has been useful. NOTE Treffsikkerhet: 89% (HØY) 00:25:39.750 --> 00:25:48.400 So, this is a way, a very strange way of using the word predict, and a strange way of using 00:25:48.400 --> 00:25:54.100 relationships among variables to answer the informal question. Does the intervention work? And 00:25:54.100 --> 00:25:59.450 that's how you go about it. And that's the sense in which the questions that statistics answer, are 00:25:59.450 --> 00:26:05.900 relationships among variables. And that's the sense in which all Sciences work with hypotheses, 00:26:05.900 --> 00:26:09.450 about relationships among variables. So basically, NOTE Treffsikkerhet: 91% (HØY) 00:26:09.450 --> 00:26:15.400 Everything we know about the structure and function of the world. Like I said, physical, biological 00:26:15.400 --> 00:26:18.650 and social domains. That's how we get it. NOTE Treffsikkerhet: 86% (HØY) 00:26:18.650 --> 00:26:24.800 And of course, because we're doing all this statistically, we don't only have knowledge about the 00:26:24.800 --> 00:26:31.900 world. We also have knowledge about the uncertainty of our knowledge about the world. So we know 00:26:31.900 --> 00:26:39.400 what we can be confident about and what we aren't so confident about yet and need to work more on. NOTE Treffsikkerhet: 85% (HØY) 00:26:40.200 --> 00:26:48.000 Let me now, go into an example of formulating, a research question, just to make sure this point is 00:26:48.000 --> 00:26:54.200 a bit clearer. So let's say you have some kind of informal question that you care about "Does coffee 00:26:54.200 --> 00:26:59.000 help studying". This is not a special education question, but it might be question that you're 00:26:59.000 --> 00:27:04.900 interested in knowing the answer because it's a student. You might be interested in knowing whether 00:27:04.900 --> 00:27:10.400 there are reasons to change your behavior. Like, are they reasons to increase your coffee intake are 00:27:10.400 --> 00:27:10.900 the reasons to NOTE Treffsikkerhet: 91% (HØY) 00:27:10.900 --> 00:27:19.000 decrease it with respect to your study outcome. This may look like a very interesting and relevant 00:27:19.000 --> 00:27:26.500 question but is unfortunately not an answerable question. And the reason it's not answerable is 00:27:26.500 --> 00:27:31.400 because it's not specified. So what is coffee mean? NOTE Treffsikkerhet: 73% (MEDIUM) 00:27:31.700 --> 00:27:41.300 Is this amount of caffeine? Is it frequency of drinking coffee? Is it espresso versus Americano? 00:27:41.300 --> 00:27:49.300 Is it the fact whether you drink any coffee at all or not? What is the meaning of coffee in this 00:27:49.300 --> 00:27:56.800 question? And this doesn't have one right answer. There are many possible answers that you can give, 00:27:56.800 --> 00:28:01.100 but this makes you think about what do I really mean by this question? NOTE Treffsikkerhet: 90% (HØY) 00:28:01.100 --> 00:28:08.800 So you have to specify what you mean, and that may even be the easy part because studying is the hard 00:28:08.800 --> 00:28:12.000 part. So what do you mean does coffee help studying? NOTE Treffsikkerhet: 91% (HØY) 00:28:12.600 --> 00:28:21.500 Does it help? Does it help you be reading a book longer without falling asleep? Does it help you 00:28:21.500 --> 00:28:23.450 get better grades? NOTE Treffsikkerhet: 91% (HØY) 00:28:23.450 --> 00:28:29.199 That's a very different questions. So what do you mean? NOTE Treffsikkerhet: 91% (HØY) 00:28:29.199 --> 00:28:36.300 If you want an answerable question, it has to be very specific. So you have to see exactly what 00:28:36.300 --> 00:28:43.000 studying means. It could be hours over a book. It could be number of books read. It could be 00:28:43.000 --> 00:28:50.300 classes that are you got a passing grade on. It could be your average grade in an exam. It could be 00:28:50.300 --> 00:28:55.350 your average grade over your entire study period. It could be a lot of things. They're not wrong, 00:28:55.350 --> 00:28:57.150 but you have to decide. NOTE Treffsikkerhet: 91% (HØY) 00:28:57.150 --> 00:29:00.300 and there's more like, NOTE Treffsikkerhet: 91% (HØY) 00:29:01.300 --> 00:29:06.400 What do you mean does coffee help studying? Do you mean that should you drink coffee during the 00:29:06.400 --> 00:29:10.950 semester? Do you mean, should you be, should you drink coffee during the exam period? Do you mean 00:29:10.950 --> 00:29:15.800 should you drink coffee right before you read something right after you read something right before 00:29:15.800 --> 00:29:22.500 you go to an exam like throughout your study period of two years. What is the meaning of the 00:29:22.500 --> 00:29:29.100 question in that respect? Well, there's lots lots and lots of details that need to be specified 00:29:29.100 --> 00:29:31.950 before this interesting informal NOTE Treffsikkerhet: 79% (HØY) 00:29:31.950 --> 00:29:36.500 question, becomes an answerable scientific question. NOTE Treffsikkerhet: 81% (HØY) 00:29:36.500 --> 00:29:43.600 And the first step is in defining variables. So variables are defined by specifying. Well, in name 00:29:43.600 --> 00:29:50.100 we can call it coffee, but the issue isn't the name is the nature in range of values. So, what 00:29:50.100 --> 00:30:00.300 coffee means number of cups per day number of cups in a semester grams of caffeine a day or what? NOTE Treffsikkerhet: 87% (HØY) 00:30:00.300 --> 00:30:06.900 And so the method by which they are assigned, do you actually ask someone? Do you measure? Do you 00:30:06.900 --> 00:30:13.600 observe? Do you give out coffee cups to people who are in this specific setting? So there are 00:30:13.600 --> 00:30:16.350 different methods to assign these values. NOTE Treffsikkerhet: 91% (HØY) 00:30:16.350 --> 00:30:23.400 And the important thing here to repeat is that you can only answer specific well-defined questions. 00:30:23.400 --> 00:30:29.600 And once you have that, then statistic can become helpful. So if we were to map out all the route 00:30:29.600 --> 00:30:36.400 from a question to an answer, we generally start with a potentially interesting but vaguely 00:30:36.400 --> 00:30:42.900 formulated informal question, and that's true of everyone is not just true of people who aren't 00:30:42.900 --> 00:30:45.850 researchers, researchers also have this NOTE Treffsikkerhet: 73% (MEDIUM) 00:30:45.850 --> 00:30:52.000 starting point. So there is an interesting question, that's too vague to be answerable and then we 00:30:52.000 --> 00:30:56.850 make this question precise. And this means NOTE Treffsikkerhet: 88% (HØY) 00:30:56.850 --> 00:31:04.400 define the relationships among specific variables and there are a lot of issues going in there. And 00:31:04.400 --> 00:31:09.400 you in this course, you're going to learn about operationalization and construct validity and issues 00:31:09.400 --> 00:31:16.400 that go in the route from what you have in mind to what you can actually measure. NOTE Treffsikkerhet: 91% (HØY) 00:31:16.700 --> 00:31:24.500 And then you measure those and you get your quantitative data, and then these data are statistically 00:31:24.500 --> 00:31:30.400 analyzed and you get an answer about the relationships among the variables and you get the 00:31:30.400 --> 00:31:37.100 confidence that's associated with the answer. And then that answer is to be interpreted and acted 00:31:37.100 --> 00:31:43.500 upon. And the thing that I want to highlight here is that NOTE Treffsikkerhet: 89% (HØY) 00:31:43.500 --> 00:31:49.050 This is where statistics comes in. So statistics is a link. NOTE Treffsikkerhet: 86% (HØY) 00:31:49.050 --> 00:31:52.949 In a chain of research methods steps. NOTE Treffsikkerhet: 91% (HØY) 00:31:52.949 --> 00:32:01.900 That concerns the analysis of quantitative data and there is a very similar chain of steps. If you 00:32:01.900 --> 00:32:07.000 were working with the qualitative method, except you would be gathering different kinds of data and 00:32:07.000 --> 00:32:12.600 you will be applying different kinds of procedures. And so you'd have different procedures for 00:32:12.600 --> 00:32:20.300 working with those kinds of data. The rest is pretty much the same. So that's why statistics is in 00:32:20.300 --> 00:32:23.250 the broader context of a research methods course. NOTE Treffsikkerhet: 86% (HØY) 00:32:23.250 --> 00:32:31.800 And forms a one-step one link in the chain of things that you have to do to go from an interesting 00:32:31.800 --> 00:32:34.600 question to a usable answer.