Okay, so if you really want the truth you have to do an experiment. we're going to talk about contrast in general, but experiments are very special kind of contrast where you create conditions with the express idea to test some idea or theory. And one of the things that experiments can get you that, for example, correlational studies cannot, is they can actually show causation and, and what we mean by that is we can actually show that a manipulation of some variable causes changes in some other variable. and that's when we really feel like we understand what's going on. So let's jump into that. All right. Last lecture for week one, knowledge by contrast. I want to just kind of walk you through an example so you get a really good sense of what an experiment is, the process, and what makes it special. So, let's say we have this theory, and this is a theory that's been around for a while. It's called evolution theory and it's some, you know, surprising to some, but it is a contentious theory, and so I want to tell you one experiment related to evolution theory, and use it as an example. So here's the theory, the, the big theory. Animals diverge into different species as a result of changes in or differences of environmental conditions, like food sources. So, Darwin said, you know, you, the survival of the fittest. By, what he meant by fittest was, those who fit the environment best survive. And the attributes that they had now tended to be the ones that were shown in, in higher numbers in, in their next generation the children that they produced. So that eventually a species will evolve to fit an environment. Okay, that's the big theory. well, can we derive a prediction, a testable prediction from that theory. Well, here's one. If this is true, then it should be possible to start with some single species of animal, but create different species by consistently exposing random sub-groups, you hope, to different environmental conditions. For example like food sources, so we can take our, our little friend drosophila here, common fly. Drosophila is the subject of many experiments. And in the one I'm going to talk about, here's what they did. They took a bunch of drosophila, a bunch of flies, and then they randomly created two subgroups. Now these subgroups were no different originally. That's the joy. That's the advantage. That's the reason why we like random assignment. If you start with a group, and then you randomly assign individuals to two different conditions, then you assume that those individuals on average are very similar. There's no reason we should expect any systematic difference, at least, across those groups when we start. Now we do some manipulation. and we want to see whether that manipulation will have an effect. On some variable. Let me come back to that, let's just do it. So let's say we have these two subgroups now that we've randomly derived from our flies. And now this subgroup we're going to give a starch type food. Not just to this group but to it's offspring and to it's offspring's offspring. So in fact We're going to allow eight generations to pass and all of those flies will only ever get access to the starchy food. OK, in contrast, this other group, which originally were the same, we're only going to let them get access to maltose food. And the same kind of idea. So not just them their offspring and their offspring's offspring, so eight generations. Of now maltose-eating drosphilia. Okay. And so the question is: Well, does this matter? If we expose them to different food sources, do we see any sense of them becoming different species? Well, here's what you do see: You will see that these starch-eating drosphilia First of all look different. They tend to become much lighter, much more yellowy, whereas the maltose food[INAUDIBLE] become much darker, much browner, a golden, sort of golden brown. so we're definitely seeing an effect of the food. It's certainly changing the physical characteristics of our research subject. And, and this is one of the things we associate with different species, right? Different species look different somehow. But there's another aspect of different species too. Different species tend not to interbreed. And so, if we now allow these drosophilia, these 2 populations to now merge and hang around. And now, we watch their mating behavior. What you will see is that they will only mate within species. So the starch-eating ones will seek out other starch-eating ones. And the maltose-eating ones will seek out other maltose-eating ones, and they will not interbreed. This is the mark of them truly, being individual species. So it certainly seems that by manipulating through source. We were able to create two species of drosophila, okay. So if you just sort of backing up a little bit, because I want to give you some of this terminology too. When we do an experiment like this, we're almost always manipulating some variable. And then looking at the effects of that manipulation on other variables. The one we manipulate we call the independent variable. And in this case, that would be the food source. So we, the experimenter, decided who was going to get which food source. So we were in control of that, and we manipulated that. OK? So that's what the independent variable is. We thou, now want to see, okay, what effects did that manipulation have? Well, on what? So what are we going to measure to look for effects? Well we want to measure species I, species I. Speciation. Thank you. We want to measure speciation and there's sort of two things relate this speciation. One is species specific markings. So we want to look at the physical characteristics for example. That would be the yellowy color versus the darker color. So according to that variable, the color of. The drasphelia you do indeed see a difference and we are also curious about this intermating because we know this is a marker of spec-, I've lost it again, my goodness. of there being 2 different species speciation there we go and so we want to look at mating behavior to see how often they mate within species or at least within food source or across food source. So those things would be what we call our dependent variables. And we hope, or we expect, that if our theory's right, and we manipulate our independent variable in a way where our theory predicts something should happen, then when we measure our dependent variable, and contrast it across those levels of the independent variable, which would be starch root versus maltose. We expect to see differences okay, so we're really looking for differences now of some depedent variable as a function of the independent variable. when we do that, the test we use is actually a variety depending on the design but the most simple test and the one we'll be playing with in the activity in this course. Is something called a t test. So I want to just give you a really general idea of what a t test is again without getting in too much of the, of the guts. But a t test is a ratio. Because what we really want to know, let's imagine the color of the drosophila. You can measure color on a continuum. And so, if this the starch-eating ones, you would see some that were you know, up up, eh, it, it well frequency of light let's say. And so let's say this is sort of a yellowy color. Well some are more or less yellow, some individual. So, most of them are probably right around this yellow color. but there's a few that are, a little on the paler side, and a few that are on the, the darker side, of the yellow. here's this other group. Whoops, I'm sorry. The maltose eating ones and we expect them generally to be darker but again, we see a distribution where we have some really dark maltose ones and some not so dark maltose ones. And we have this whole distribution. And what we want to know is, on average. Did the starch eating ones seem different than the maltos eating ones? Is the average starch-eating drosophila more light than the average maltose-eating drosophila? So, is this difference, something real, is it, is there a difference? Well, that's a tricky question to ask, it turns out, and this is where we bring in statistics, and a branch of statistics called inferential statistics. Generally though, what it does is it says well, first of all I want to know how different these two groups are. so we have our, you know, yellow ones and our darker ones. Well how different are they? And it When it's trying to decide if that difference is different enough it also wants to know well, how different are individuals within each of these groups. What's the variability within the groups? And so that'll give me some notion of how much things vary just by chance when we don't manipulate anything. And now we can look at the difference caused by the manipulation and this gives us an idea of how big that is. Is, if this is a lot bigger than we would expect by chance, then we're going to get excited and say we got something here. But if it's not much bigger than we expect by chance, then we don't get that excited. So we compute this t value. And large t values. That is T values that is different from zero. the further different from zero the more excited we get about them. That means there's something going on. And so literally, what we would do in an experiment like this is compute a T-value based on the data we have, and then we would compare it to some critical value, and this critical value would tell us What size of T is big enough for us to get excited? And that'll depend on the, the number of subjects we and, and a variety of other things that we're not going to get into in detail here. But the general process would be to have two groups, get some variable, like color where you measure the average for each group, and you get a sense of a variability within each group. You then compute a t value, and you compare that t value to the critical value. And if your t is more extreme, I'm being hedgey here because t's can be positive or negative depending on the order that you subtract. And really the positive or negative isn't important. It's how different from zero it is So if a t is more extreme, bigger than the critical t, then it's real. And we get excited, and we say hey, we found something. Okay. You're going to get to do some of this. This is partly why I'm taking the time here. Because you'll be involved in this process. And that'll give you a really good sense of it. Alright? >> So when you get a result like that, you say you have a significant result and that's very exciting. Your experiment is a success. But there's other, if we think in the bigger picture of what defines good science, there are other factors. and you're going to play with all of these factors in the digital lab code activity. So one is, well it's really great that you found some difference. But is that difference interesting? Is it really interesting, is it really relevant. and this is really left up to the scientists, to really make that case. And, and what I want to highlight here is part of good science is good marketing. Those scientists who can express to others why there data is important often get the headlines, and other scientists who may have just as important findings but don't market it as well can sometimes fall by the wayside. Nobody reads their papers. So indeed marketing is part of science and marketing should be based on how interesting or relevant Some result is. Another thing that's very important is the notion of what we call replicability. And what this means is, okay, you went out and you did your experiment say on some drosophila and you found a certain result. Cool, what we'd really like is if somebody else now redoes your experiment and finds the same results. If that happens with a different set of flies. And they do it all over again and they get the same thing. Then we say they replicated your experiment. And that makes us feel really good that we can trust your experiment, so replicability is nice. What we then really hope to see is generalizability. So what we mean by this is, okay, it seems like the evolution theory, at least the prediction derived from it, worked on drosphilia. Would it work on snakes? Would it work on you know, various other kinds of animals we try? So we might want to actually try it on spiders or on something else. Do the same kind of experiment, generally, and if we see the same result When we change aspects of the experiment, like the population, or like how we measure things, then that makes us feel like this is a more general finding. Something we can believe to be a general truth. That yes, specia, specia, speciation, does seem to reflect changes in environmental conditions. so these are other things and you'll be able to play with some of these in that project. Okay, so that's as much science as I'm going to nail you with, but I know right now a lot of this stuff is ambiguous, but wait until you get your hands dirty. Once you get your hands dirty, you'll start to see that this isn't only understandable, it's kind of fun. and, and that's what Digital lab code is all about. It's about making it fun. Here just, I, I have a couple of videos that are really about experimental design and the scientific method just to echo and in fact you know, go beyond some of what I've told you. and now here are a couple of readings about T-tests again, showing you a little bit of the math if, if you're interested, and the general notion of comparing two different groups. So check those out and then take what you learned and go do that digital lab code activity. have fun doing it. And at the end of that, I think when you reflect upon what you learned, you'll realize that you are now a scientist. Congratulations.
