st095 ยป

st095 Lesson1

Contents

What would you do

Welcome to Lesson 1. We're really excited to have you taking this class and we hope you like it. We're going to jump right in with a poll. And don't worry, this doesn't affect your grade. We just want to see what you're going to say. So, you have, have big exam tomorrow, let's pretend, and your memory needs to be as sharp as possible. What would you do? Would you stay up studying? Get a good night's sleep? Refrain from partying? Get a good workout? Or eat a good dinner?

Believe results

What if I told you that I did a survey and would get a good workout, 28% said they would eat a good dinner, 16% said they would stay up studying, 13% said that they would get a good night sleep, and 1% said that they would refrain from partying. Would you believe me? What would you want to know in order to really believe these results? How many people I surveyed? Who I surveyed? Or how the survey was conducted?

Believe results

You would want to know all of these things. Each of these are factors that influences the validity of our results. A good sample size, a representative sample, and methodologically sound research are all really important in a good research study. When you read research, you should always be aware of these things. That way, you can make good decisions based off of valid research.

Measure memory

Let's assume that to get a good grade, we need a good memory. That means that we have to do everything we can to boost our memory. Therefore, we might look at a bunch of different factors that influence memory. If we're doing a real research study, we want to make sure that these factors really do positively influence memory, but we can't analyze the relationship between these and memory without having a clear definition and way of measuring memory. So, how would you measure memory? There is no right or wrong answer, the purpose of this is just to get you thinking.

Define constructs

How would you define happiness? being at peace. The ability to [LAUGH] I guess live a fulfilled life. it's more of what you feel, what you can bring to others and not so much what you see. They kind of bounce, their eyes get all happy, their ears kind of go up, they get all loose like their body language is real loose. and yeah their tails wag, like he's happy right now. How would you measure it? Oh, I don't know. I don't measure nothing really. I just, I'm happy. I think I'd probably measure it by the number of times I smile every day. I have no idea. I'd, I'd have to think about that for a sec. Their, their attitude, their, just the way they are personally. If, if I was a researcher and I wanted to know how well you remembered something. How could I do that? if you're taking about like, say scientific experimentation, if you have a control of some sort, you could film the experience that a person has and then interview that person. maybe have multiple people to interview, so you get different points of view for the same event and see how that differs. How would you define memory? it's an accumulation of information, data far more important is how you apply that, how you use it. the ability to retrieve information from your past. I have to think about it. Things I bother to remember for later use. How would you define itchiness? Discomfort, yeah, just, just general discomfort. Something that's irritating you and you have to, you have to scratch it. The intensity of the scratch would match the time length of scratching. with the amount, that they move around. You could ask a very forward question like, is there something you would need to, you need me to scratch, right now? Well, that sounds a little personal. very close, very close friend, maybe you could do that. So kind of like happiness, if you had to measure how itchy your dog was right now, how would you do that? How would you quantify? I'd, I'd definitely go with like how often they were scratching or how often they were doing sort of a full body shak e. How would you measure stress? I think you could measure stress by, obviously, there's I'm sure medical ways that you could measure stress. Yeah, or the frown on my face, on my, on my, mouth, I mean. [LAUGH] Heartbeats probably. The blood rushing to the head.

Clearly there isn't just one way to measure things like memory, happiness, guilt, love and these are called constructs. It's very difficult sometimes to define and measure them, and so there are a lot of different ways that we can do that. BBC came up with one way of measuring memory, by creating a test on how well you can remember faces. Click the link at the right called BBC Face Memory Test, to do this memory test. Remember your scores or write them down. And again, this isn't graded. We're just going to do this because this is a formal way of measuring and testing memory.

Great, so you should have gotten both a recognition score and a temporal memory score. The recognition score measures how well you recognize faces that you saw before and the temporal memory score measures how well you remember when you saw each face. We're going to use your data to analyze the construct memory. Click the link at the right called BBC Scores to enter your scores from the Memory Test. And don't worry, we don't really care what your scores are and we still think you're awesome. If you got 68% for example, just type 6, 8. You'll also notice that we ask how long you slept last night, so please enter that as well.

Throughout this course, we're going to be sharing a lot of data with you. Some of it will be data we collect, maybe from you guys. And some of it might be data that we find. Either way, we're going to try to make it really fun and interesting. And you'll learn a lot in the process. We think that the easiest way to share this data with you is with a Google account. If you haven't already, pause this video now and create your Google account so we can share this data with you. This will be really important throughout the entire course.

BBC Measurement

So now, we're going to take a quiz. And this is a little tricky, but, of course, the whole point is to help you out, get you thinking about measurement. So how did BBC measure memory? Note that we're not asking about factors that contributed to your score. We're asking specifically how did BBC precisely measure memory. Was it the types of faces you remember, the percent of faces correctly recognized and placed from parts I and II, knowing whether you saw the face first or second, knowing whether or not the face was there, Or the number of faces you remember?

BBC Measurement

So the correct answer is this one. The percent of faces correctly recognized and placed from parts I and II. So this is very precise, the percent, of faces. The others contribute to your score. So, if you remember specific types of faces that helps you to get a, higher percentage. Or knowing whether you saw the face first or second. That also helps you get a better score. But that's not how it was measured. Knowing whether or not the face was there. Yes, that contributes to your score. And the tricky one. The number of faces you remember. If the score were given to you as a number. For example, you remembered, I don't know, 11 out of 12. That would be the way of, that BBC measurement but your score at the end was a percent. So this is how they measure memory, and it's important to be very precise. And if you got that wrong, don't worry. This was a tricky question and it's more just to get you thinking.

Operational Definition

Now, when we pick a way of measuring it, we also kind of have an operational definition. So in the BBC study, the operational definition of a good memory is when the percentage of faces correctly recognized and placed is higher. So once we have an operational definition, we're able to measure constructs in the real world.

Constructs

Now as we said before, constructs are very difficult to define and measure, and everyone might have their own definition and way of measuring it. So now we're going to take a quiz to see how well you understand constructs. Choose all that you think are constructs.

Constructs

Gallons of gasoline, is not a construct, because we already have a way of measuring it in gallons. Intelligence is a construct. How do we measure and define intelligence? We have the IQ test or maybe, we could measure intelligence with grades. But grades may also help us measure something like how hard we're working, like effort. Effort is a construct. We might measure effort by how many minutes we spend working on homework, or by grades, or GPA. Age is also a construct. If I had put age in years, I would have defined age as the number of years you've been living, but some people might look at age as how mature you are. Here, I didn't put a way of measuring it and age therefore, is kind of tricky, would be considered a construct in this case. So, maturity, age in years, and wisdom are all three different ways you could possibly measure age. Hunger is also a construct. Hunger could be measured by how often your tummy grumbles or by how deficient in nutrients your body is. Annual salary in US dollars is not a construct. That's because we put in US dollars and therefore, we have a way of measuring your annual salary. And finally, itchiness is a construct. How would you define and measure itchiness? I'll let you ponder that one yourself. We have to think carefully about how to define and measure constructs. The descriptions for constructs that we settle on and that allow us to measure them are called our operational definition.

Operational Definitions

So, try and match each construct, here, with a possible operational definition. Now, these aren't necessarily the operational definitions for each of these constructs. These are just examples of how we could measure them.

Operational Definitions

Good job trying to match each construct with each operational definition. For depression, someone's score on the Beck's Depression Inventory could be a way of measuring and defining depression. Fur hunger, grams of food consumed. For stress, levels of cortisol, which is the stress hormone. For anger, this could either be number of profanities uttered per minute or even ratio of minutes spent smiling to minutes not smiling. The difference is that the ratio of minutes spent smiling to minutes not smiling would be a better measure of happiness, because if its greater than one, then you're spending more time smiling. So, the higher this ratio, the happier you would be. So, this would be a better measure of happiness. So, for anger, a better operational definition would be the number of profanities uttered per minute. Some operational definitions aren't so good, but this is just an example. Health could be resting heart rate. Obesity, body mass index. Effort, minutes spent studying for an exam, and brand loyalty, the number of products purchased per year from a particular brand. So these are all operational definitions which allow you to measure each of these constructs. And after being able to measure these constructs, we can analyse them

Data

Data is the most essential part of statistics. Without data we couldn't do anything. So now we have data on how many hours you slept, your recognition score, and your temperal memory score. And we have data values for each one of you who enetered these scores. So the first person slept 7 hours, got a 91% Recognition Score, and an 86% Temporal Memory Score. So this row of data is for one person. The next person got 6.5 hours of sleep, a 95% Recognition Score and a 78% Temporal Memory Score and remember that you can see this actual data if you click on the BBC results link. Anyway, so we have columns of data and each row corresponds to one person. Hours slept, Recognition Score, and Temporal Memory Score are called variables because they vary across individuals. So now we're returning to the question. You have an exam tomorrow. Should you get a good night's sleep tonight? Does it matter?

Sleep and Memory

What if we looked at our data and noticed that everyone who got less than six hour of sleep also got a temporal memory score less than 70%,, and that everyone who got greater than six hours of sleep, got a temporal memory score greater than 70%? So that would mean their scores are either here or here. Which of the following conclusions is the most likely? The less sleep you have, the better your memory. The more you sleep, the better your memory. People always get the same memory score no matter how long you slept or there is no relationship between sleep and memory.

Sleep and Memory

The answer, if this were true, is that the more you sleep the better your memory. People that didn't sleep for very long got lower scores. People that sleep for longer, more than 6 hours, got better scores. So if all the date followed this pattern, we could reasonably conclude that the more you sleep, the better your memory tends to be.

Influence Memory

Besides sleep, other things influence memory. This is supposed to be a brain. So, what else do you think could have influenced your memory? The time of day at which you took the test, the number of palm trees in Hawaii, whether or not you're stressed, the price of milk, whether or not you took the 5 minute breaks that the test recommended, the mass of Pluto, your age and finally, not paying attention when their faces are shown. Check all that apply.

Influence Memory

So definitely the time of day that you took the test could have influenced your memory. If you're taking the test late at night you might be really tired, or maybe your brain just doesn't work in the morning like one of my friends. He can't think at all in the morning. That isn't to say that it definitely did influence your memory, but it could have. The number of palms trees in Hawaii isn't very plausible to have influenced your memory. Your stress level definitely could have influenced your memory. If you're really stressed, maybe you don't remember as well. The price of milk not so plausable. The 5 minute breaks, absolutely. If you didn't take the 5 minute breaks, my hunch is that it might be more difficult to remember which face was from part 1 or part 2, the mass of Pluto, not so much. Your age? Definitely, and finally, not paying attention definitely could've influence your memory. The BBC study concluded that age and time of day Were the two biggest influencers on how well people performed on this memory test, if you didn't get a chance to see the BBC test results. So we're looking at a bunch of different things that could have influenced your memory, including sleep. We make hypotheses all the time, every day.

Control for time of day

So as a reminder, we're specifically concerned about whether or not sleep influences memory, but we've identified all these other variables that could influence memory. Now it's time for a quiz. So would you trust this data more If you knew that everyone had taken this test at the same time of day, meaning the same time of day for everyone's respective time zones. So for example someone in India took it at 5:00 their time, same as someone in California who took it at 5:00 their time. Yes or no?

Control for time of day

Yes, we would trust the results more, because we're holding one of these factors constant. When more factors are held constant, the test is more reliable because these conditions are the same for all people that took the test. We always have to think about extraneous factors. Things that can impact the outcome that we may not have thought about. These extraneous factors can impact the outcome, in this case, our memory. These extraneous factors are often called lurking variables. They just lurk around waiting to ambush your results. It's really difficult to account for every possible extraneous factor, but this is something to be aware of not only as you're going through this course, but when you analyze data in your daily lives.

Same scores

You may have seen after taking the BBC Face Memory Test that the average recognition score of everyone they studied was 92% and the average Temporal Memory Score was 68%. We're going to ask you a quiz question, and remember this is just to get you thinking. Do you think that the average scores for this class will be exactly the same as these averages? Yes or no? And when you think about this, think about why you would say yes and why you would say no.

Same scores

Now the answer is no. Within the population of everyone that took the BBC test, you guys are a sample. There are a lot of different samples we could take of all the people that took the test. We can take big samples, and small samples, and while your sample could get the same averages as those from the entire population, There will also be samples that don't have the same averages. The average memory score of the population is called the population parameter and it's denoted by Mu. It's a funny little Greek symbol. And the average for the sample is called the sample statistic, and we generally denote it by x bar, an x with a bar on top. Now, I want to stress, please don't be nervous about symbols. I know a lot of people get pretty nervous when they see symbols that they haven't seen before, but, in mathematics, we really need symbols to illustrate our points. Think of it like, when you're playing a musical instrument, and you have sheet music. How else are we going to make sure that people in future generations learn these classical melodies? Sheet music is the best way to do that, and same with mathematics when we use symbols. We need to be able to articulate our thoughts to other people all around the world and that's one thing that's really cool about mathematics. It's a universal language. So, don't worry, you'll soon become familiar with these symbols, because we're going to use them throughout this course. Now, we can use this sample statistic to approximate the population parameter. But, generally they won't be exactly the same. The difference between these 2 is called the sampling error, which means we can make educated guesses about population parameters using sample statistics but we probably won't be one hundred percent accurate. That's why it's really important to have a good sample that can better predict the population sample. So going back to the quiz, no, we don't think that the average scores for this class will be exactly the same, but it'll probably come close, and we're going to be able to se e that later.

Sample average

So we're going to spend a little more time with population parameters and sample statistics. Let's say that we have these memory scores. And what if I pick a sample from this population that consisted of these scores? Now, just a point of clarification. Populations are generally much bigger than this. But we're just going to pretend that this is the population. And we're choosing these 3 scores for a sample. Will the average of the sample be greater than, equal to, or less than the population average? What do you think?

Sample average

So, without even needing to calculate anything, you can see that we chose the biggest scores to be in the sample. All the other scores are a lot lower. And so therefore, we automatically know that the average of this sample is going to be bigger than the population parameter. So this is just an example where sometimes we'll take a sample and it won't accurately estimate the population parameter.

Better sample

So now we're going to have another quiz. What will make the sample statistic x bar closer to the population parameter mu? A smaller sample, or a bigger sample? What do you think?

Better sample

So, the answer is a bigger sample. If we choose this one also and lets say, these see we have this small value 32% and that's going to drag down the sample statistic, x bar, the average. So, this will be closer to the population average, mu. And, of course, ideally, we want the entire population but oftentimes, that's really hard to do because the population might be so big. That's when we need a sample. But a bigger sample will better approximate the population parameter. Another example where samples can be misleading is predicting the outcome of the Presidential election. For this last Presidential election with President Obama versus Mitt Romney, a lot of polls were given out which each came up with its own result. Some of them predicted that Romney was going to win which, as we know, didn't happen. While each sample may not perfectly predict population parameters, the sample statistic can give us an interval in which the population parameter lies as long as this sample is random and as unbiased as possible. We'll go more into depth about samples later.

Random selection

We're going to take one more quiz just to get you thinking about random samples. Let's say that there are n people or objects, whatever, in this population. Which of the following samples are random? So one option is to assign a unique number, from 1 to n, to each object in the population, then randomly choose numbers from 1 to n. So, let's say there are 10 objects in the population. So we assign each of them a number, 1, 2, 3, 4, 5, 6, number generator to select a few random numbers between 1 and 10. Or, another option, assign a unique number from 1 to n, just as we did before, and then choose all multiples of 10 to be in the sample. So, let's say there are 100 in the population. Then, the object that got assigned to number 10, 20, 30, 40, etcetera, those will all be in the sample. This one is pretty similar to this option, exept we randomly assign a unique number, multiples of 6 to be in the sample. And finally, your option is, randomly assign each object in the population to 1 of 2 groups, so we split the population in half and then randomly select a sample from each group. Check all that you think will result in a random sample.

Random selection

So the first one is random, we assign a unique number 1 to n to each object, and then we randomly choose numbers from 1 to n. This results in every object in the population having an equal chance of being selected. The second one's kind of tricky, we don't know for sure if it's random, because we don't know if we randomly assigned a unique number. We have to make sure that if we're going to give each object a number it has to be randomly assigned and each object has the same chance of getting any number. This one, we took care of that. We randomly assigned a unique number and then we chose multiples of 6. Now this choosing multiples of 6 might throw you off a little bit because that doesn't seem random. But every object in the population has an equal chance of being a multiple of 6. And finally, this one is also random. This one's actually called stratified sampling because we divided the population into sub-populations in this case 2. And every object in the population had an equal chance of being in either one of the sub-populations. And then, random samples were taking from each group. so this is also random. So yay. You're done with lesson one, moving on to the problem set.

Relationship

So, take a look at this data. Here, we have the variable hours slept, and the variable temporal memory score. So, which of the following do you think best explains the relationship between these variable?

Relationship

You may have noticed that it's very difficult to judge patterns based off of just lists of data. So now, let's explore a better way.

Visualize relationship

This is a scatter plot visualizing the data in this table. Each point represents one row of the table. Here you can see that the hours slept was about 5 and the temporal memory score looked to be about the x-axis and temporal memory score is on the y-axis. We call the variable on the x-axis the independent variable or the predictor variable. And we call the variable on the y-axis the dependent variable or the outcome. We're trying to predict temporal memory score using hours slept. Now that you've seen this data visualized, is there a relationship between hours slept and temporal memory score? The more you sleep, the better your temporal memory score. The more you sleep, the better you'll do on a test. The more you sleep, the worse your memory or there is no relationship.

Visualize relationship

It's a lot easier to visualize this data with a scatter plot. And if we look at all the values, and draw an oval around them, it looks like there is a clear upward trend. This provides evidence in support of a higher temporal memory score with more sleep. However, we don't know that a higher temporal memory score will also result in a better test score. So we can't say this. This third option is the opposite of the trend that we can see here, and finally, there does look like there's a clear relationship between hours slept and temporal memory score. People who slept longer tended to have higher scores. People who slept less hours tended to have lower scores. When we visualize data, it's usually a lot easier to draw conclusions.

True or not

Well, I showed my friend this data. And now, he thinks if he goes to bed early, his memory will definitely be better tomorrow. Is this necessarily true?

True or not

No, this is not neccesarily true. Look at these 2 people. They slept different amount of sleep, 6 hours and 8 hours, but got the same temporal memory score, about variables or extraneous factors? We haven't taken all of these into account. In other words, we haven't controlled for lurking variables. Sleep could influence memory and by this trend it looks like it probably does. But many other factors can also influence memory, and this varies by person. There's a relationship here between hours slept and temporal memory score, but that doesn't mean that hours slept causes a higher temporal memory score. In other words, correlation does not prove causation.

Golden Arches Theory

Now, this is a really important concept and we're going to try to drive this point home. Thomas Friedman said in his book, The Lexus and the Olive Tree, that no two countries with a McDonald's have ever gone to war since opening the McDonald's. This is called the Golden Arches Theory of Conflict Prevention. Let's just assume that after hearing Thomas Friedman's observation, policymakers concluded that if every country built at least one McDonald's, we would attain world peace. Well, what could be a plausible explanation for Friedman's observation and which would lend for different policy conclusions? Let's ask some people what they think.

McDonalds = World Peace

Well, what do you think? Which of the following could be a plausible explanation for Friedman's observation, and which would lend for a different policy conclusion? One option is this is completely plausible. McDonald's makes people happy and happy people don't go to war, or maybe countries with McDonald's spent too much of their money in order to open the McDonald's and now they can't afford to go to war. Or maybe citizens of countries with McDonald's are too unhealthy to go to war. Or maybe countries with McDonald's are more open to globalization and foreign investments and less inclined to go to war with other such open countries.

McDonalds = World Peace

It sounds logical that happy people don't go to war. But we don't know that McDonald's makes people happy, for one thing. And even if it does, many other factors could make people unhappy again. For this one, I'm thinking that countries that have McDonald's tend to be more economically developed and therefore have more money. So it's implausible they wouldn't be able to afford to go to war. That's my own hunch, but I'm pretty sure this is implausible. For this one, we don't know anything about how many people eat at McDonald's and how often, and we don't even know for sure if eating McDonald's would make people too unhealthy to engage in a war. This one seems to be the most plausible option. Countries with Mc Donalds are open to globalization and are less inclined to go to war. But I have to admit that would be pretty awesome, if we could obtain world peace just by building a Mc Donalds in every country. If that were true I would definitely advocate for that.

Causal inference

So now, you've seen that even if we can observe a pattern between 2 variables we always have to consider lurking variables. Referring back to our sleep data, we've seen it there's a relationship between hours slept and temporal memory, but we can't necessarily be confident that sleep causes better memory. What if we wanted to prove that sleep causes better memory? How could we do this? One option is to reverse these two. so analyze hours slept as the dependent variable and temporal memory score is the independent variable. Another option might be to look at the relationship between sleep and forgetfulness. Another option is just to ask people if their memory's better when they sleep more. And a fourth option is to do a controlled experiment.

Causal inference

These first two options are again observational studies, and they can't tell us causation. If hours slept were the dependent variable and Temporal Memory Score were the independent memory variable, we would get the same type of relationship. It would pretty much be the same exact graph just reflected across the line y=x. And if we looked at the relationship between sleep and forgetfulness, we'd again be doing an observational study, but we'd probably find a negative relationship. For example, if this were forgetfulness instead of temporal memory score, we might find that people who didn't sleep that much had high forgetfulness levels. And people who slept a lot have low forgetfulness levels, meaning they remember things better. But again, this just shows a correlation, not causation. These last two, however, could provide some help.

Benefits of Surveys

Surveys are used a lot in social and behavioral science. In fact, I used the eduacation longitudnal study for my masters thesis. I looked at a lot of different factors that might influence student effort in math class. What do you think are some benifets of using a survey to do research? It's one of the easiest ways to get info on a population. They are relatively inexpensive. They can be conducted remotely or anyone can access and analyze survey results. What do you think?

Benefits of Surveys

Surveys are one of the easiest ways to get info on a population. Sometimes it is really difficult to conduct the survey, but surveys are often a lot easier than doing, say, a controlled experiment. They're also relatively inexpensive. They mostly just require time to implement and for the respondents to write their answers. Surveys can also be conducted remotely. You might have gotten a survey in the mail at one point. And finally, this is my favorite thing about surveys, anyone can access and analyze survey results if the survey owner is willing to give out that data, but assuming they are, the survey results are timeless. The National Center for Education Statistics created this education longitudinal study and thanks to them I was able to look at factors that influence effort, including how much students enjoy school, or how much students value school, or their relationship with their teachers, their gender, and a bunch of other things. It was really fun and interesting to analyze this data. Looking at someone else's survey results can be a lot easier than conducting your own survey.

Downsides of Surveys

But just like with any research, surveys are prone to all kinds of problems. What do you think are some downsides to surverys? Untruthful responses, biased responses, respondents not understanding the question, or respondents refusing to answer? There can be more than one downside.

Downsides of Surveys

All of these are downsides to surveys. We can't always be sure that respondents are answering truthfully. Even if surveys are anonymous, people don't always give the truth for some reason. It's probably a psychological thing. And oftentimes without realizing it, we give biased answers. Everyone has their own bias based on where they grew up, the experiences that they went through, their beliefs and values. When respondents don't understand the questions, we get what's called response bias. That's why it's very important for those who write the survey to be very precise in their wording. And finally, some people just refused to answer. They don't feel comfortable answering that or they're incapable of doing so. There might be a certain group of people who don't answer the questions. And therefore, the data that we get from the survey doesn't accurately represent them. When certain people refuse to answer, this is called non-response bias. Remember when we talked about constructs? Surveys are often used to analyze constructs. And like I told you, I analyzed the construct effort. As you've seen, there isn't just one definition of a construct, there are many. Therefore, it's crucial to use as subjective measurements as possible. Surveys really have to be carefully thought out and worded.

Placebo

Besides surveys, controlled experiments are another important type of research. Imagine that researchers are testing the effects of some sleep medication on a bunch of different people, and that these people are a random sample. This is as opposed to a convenient sample in which people who are convenient to find are used in the study. Everyone recieves a pill, but for some, the pill has medication that supposedly helps you sleep, and for others its inactive. The pill looks the same for everybody and the sample is randomly assigned to one of the two types of pills. What do you think the purpose was for giving some of the people an inactive pill? To make sure there are no side effects of the active pill, to have a comparison group to those who took the active pill, or to see if the inactive pill can help people sleep. Which do you think the purpose was?

Placebo

This answer choice is correct. There are two groups of people. And so that way, researchers can compare how well the active pill helped people sleep compared to the inactive pill. In this particular experiment, we can't really test for side effects of the active pill. And also in this experiment, researchers aren't really concerned if the inactive pill can help people to sleep. To figure this out, they would have to compare the effects of the inactive pill to something else. But the purpose is to find out how well the active pill will help people sleep.

Blind

Furthermore. participants are not told whether or not they received the active pill or the inactive pill. Why do you think this is? Is it because deception is important in all good research, because they may not participate if they knew they were receiving a drug? To make them all believe that they're receiving medication or they may not participate if they knew they weren't receiving a drug?

Blind

The researchers want all participants to believe that they're taking the medication. This prevents participants from being biased concerning the effectiveness of the drug. If they all believe that they're taking the medication, then we're controlling for this bias. Not letting the participants know which treatment they receive is called blinding. Researchers use these techniques when they suspect that knowing which treatment you're receiving might influence your behavior. The inactive pill in this case is called the placebo. Placebos are fake treatments so that the control group is unaware that they are being treated differently. When I say they are treated differently, I mean they're not receiving the medication. The researchers didn't tell the participants which pill they recieved because of something called the placebo effect. People who know they're taking a placebo might subconsciously think the placebo is not doing anything, and this will influence the results when they report how well they slept. Getting back to the quiz answers, this first option is untrue. Researchers only used deception when they suspect that knowing the actual intent of a study may influence the participants behavior. And in that case, information is kept hidden until the end of the study. These two options are also not true. Participants in experiments need to provide their informed consent before the research can expose them to any treatment. Letting participants know that they may end up in the control group is part of the informed consent process.

Double blind

To continue with our story, these participants are randomly assigned to either the control group, where they take the placebo pill, or the experimental group, where they take the active pill. They take the pill at the same time of day, on the same day and the same place, and then they all go to bed in a sleep lab at the same time. While sleeping, researchers observed the participants sleep patterns, and rate their quality of sleep. Do you think these researches should or should not know which treatment participants received? Why or why not? Yes, because their ratings will depend on which treatment the people received? No, to help maintain participant confidentiality? Yes, because the researchers' ratings will be more accurate? Or no, because if they know, their judgments may be biased? What do you think?

Double blind

In this case it's better if researchers do not know which treatments the participants received. This is called a double-blind experiment because neither the participants nor the researchers studying them know which treatment they received. Single blind is when only the participants are unaware of their treatment condition. And the reason this experiment is double blind is because if the researcher knows which treatment each person received, that may impact their judgements on how well the active pill works versus the Placebo. It may subconsciously cause the researchers to think that the people who took the active pill are sleeping better. And this might consciously or unconsciously alter their measure.

Controlled factors

Let's take one more quick quiz to wrap up our experimental been yet. What are all the factors we controlled for in this experiment? You may have to replay a few of the videos if you don't remember, but that's okay. Check all that apply.

Controlled factors

We controlled for the time at which participants took the pill, the place at which they slept because they all slept at the sleep lab, and what the pill looks like. It looked the same for both the active and the placebo. We did not, however, control for gender and age. We can however minimize the impact of these variables on the results. Let me elaborate on this a little bit. Let's say that only females were assigned to the active and only males were assigned to the inactive pill. Then if people who took the active pill slept better, we wouldn't know if it was due to the medication in the pill or maybe women just sleep better than men. In this particular experiment, we're not controlling for these demographic variables, but we can minimize their impact on our results through random assignment.

Random assignment

This is why participants should be randomly assigned to treatment. Meaning, everyone has an equal chance of being assigned to either of the two treatments. In theory, randomization makes the two groups similar, on average, for example, the groups should be similar in age, the number of males and females, their sleep habits, etc. And that's why it's good to have a big sample size. Randomization works best with larger groups. We'll come back to the benefits of larger samples later in this class. Now, one more thing before we end this story. We started this module with the question, if you're preparing for your exam tomorrow, should you stay up studying or should you go to sleep? Researchers have demonstrated using both observational and experimental studies that sleep is important for memory. Getting a good night's sleep will likely pay off more than a couple extra hours of studying. On the right, you'll find a link to an interesting story if you'd like to learn more about this topic. I wish I had have known that in college.

Control for what

You may have already thought of this but another thing we could do is two memory tests on each person. One where they don't get much sleep and on where they get a lot of sleep. Then we could compare the results between each persons test. If we did this what would we be controlling for? Would we be controlling for differences in memory capabilities due to gender? Or differing amounts of sleep amongst individuals? Or the variation in people's individual memory capabilities? Or time of day at which subjects took the memory test?

Control for what

We're not controlling for differences in memory capabilities due to gender in this particular test. To do this, we could divide the group into males and females and then see how sleep influences memory in both groups. We want the amount of sleep to vary amongst individuals so that we can see how their memory is related to that. We are controlling for variation in people's individual memory capabilities, because by using this same person twice, we're doing what's called a within subject design. That way we can see how sleep influences memory for this same person. And by using the same person twice, we're keeping a lot of other things constant, including this variation in people's individual memory capabilities. We're also not controlling for the time of day in this study, but this would be a good thing to add on to the within subject study. We could tell our subjects to sleep a lot, then take the memory test at the same time the next day as everyone else, like three o'clock. Then again, we can tell people to sleep a little, and then take the memory test at the same time of day again, at three o'clock.

Katies hand

Now, you've seen observational studies where we correlated sleep and memory. And you've learned about surveys where we ask people questions. And you've learned about experiments, which are our best bet for proving causality. For the most part in this course, we're going to focus on observational studies. That's because we can't exactly do a survey or an experiment online. But we do think it's important for you to know the different kinds of studies that we use with statistics. Here at Udacity, we did our own little observational study, which we're going to share with you, so that you can draw your own conclusions. By now, you know my hand pretty well. You see it all the time when I'm writing out the lessons. So now you're going to use what you know about my hand to know a little more about me. There's a link at the right called height and hand length, with real data on hand length and height of people here udacity. Let's bring this data up. If we highlight the data, and then go up here to insert chart. Go to charts and select scatter, and we need to click here on scatter chart, and then click insert. After you click insert a scatter plot should pop up that visualizes this data. You should see a clear relationship between height and hand length. Based on this scatter plot, if I tell you my hand is 6.75 inches long. How tall do you think I am? Write your answer in inches since that's how we measured hand length. Just take a guess, this doesn't affect your grade. The purpose of this is for you to use patterns to make educated guesses.

Katies hand

For the answer, we decided on a certain interval in which my height is feasible. If you guessed within this interval, congratulations. It seems like you understand the pattern shown in this scatter plot pretty well.

Draw conclusions

Well, this first one doesn't work, it's not really feasible, because your nails grow independent of how long your hand is. We also can't asset this second one. We can't simply say that taller people have longer hands. In fact, if we look at the data, these two people look like they have about the same height, but they have very different hand lengths. Therefore, we can't say with certainty that taller people automatically have longer hands, there's just a trend in that direction. Also, if we look at this trend, we see that people who have about my hand length, which is 6.75, tend to have heights around this area. Seven feet is 84 inches, which is way up here. Therefore, it's very unlikely that I'm taller than seven feet, but I could be. Remember, this relationship shows a correlation, not a causation, therefore, we can't draw this conclusion. However, we can say with certainty that people who have longer hands tend to be taller.