Latest / The Psychology Undergrad Podcast / Multiple Regression, Mediation & Moderation: Complex Correlational Designs
Transcript
- 0:00Imagine scrolling through your phone, right,
- 0:02and a headline practically screams at you. Something
- 0:05like, um, coffee causes happiness. Or maybe,
- 0:09you know, wealth makes you a terrible person.
- 0:12Right. And you just instinctively want to believe
- 0:14them. Exactly. But like, how do you actually
- 0:16know if it's true or just some statistical illusion?
- 0:19Welcome to the psychology underground. It's basically
- 0:22the ultimate question, you know, because we're
- 0:25constantly bombarded with these wild claims about
- 0:28human behavior. Oh, yeah. Every single day. And
- 0:31usually the actual truth is hiding behind some
- 0:34Well, some very complex math. Which is exactly
- 0:37our mission for this deep dive today. We are
- 0:39pulling out our class notes to decode how psychologists
- 0:42make sense of messy real -world human behavior.
- 0:46Because, you know, they can't just lock people
- 0:48in a pristine lab to run a perfect experiment.
- 0:51Yeah, unfortunately they can't. Right. And for
- 0:53you listening right now, our fellow learner in
- 0:55this study session, by the end of this deep dive,
- 0:57you are going to understand the hidden math behind
- 1:00why we act the way we do. yet really gives you
- 1:03the tools to look past the clickbait. Because
- 1:06most of those flashy headlines, they're actually
- 1:08pulling from correlational research. Which is
- 1:10quite different from an experiment. Exactly.
- 1:13And understanding how that correlational research
- 1:15actually works is really the only way to know
- 1:18if a claim is legit. So let's start with the
- 1:21most obvious question here. We always hear about
- 1:23experiments in psychology. The gold standard.
- 1:27Oh, absolutely. The holy grail. Yeah, you manipulate
- 1:30one thing, you measure another, and boom, you
- 1:31have cause and effect. Why don't researchers
- 1:34just run an experiment for everything? Well,
- 1:37because the real world is incredibly stubborn.
- 1:39Yeah. And thankfully, we have ethics boards.
- 1:43Right, yeah. Small detail. Yeah, just a minor
- 1:45thing. I mean, you can't ethically force a group
- 1:47of people to, say, experience a traumatic event
- 1:51just to see how it alters their memory five years
- 1:53later. No, definitely not. And you can't artificially
- 1:57assign a child to be raised in severe poverty
- 1:59just to measure the impact on their education.
- 2:02There are massive, you know, practical and ethical
- 2:05walls that prevent us from manipulating the variables
- 2:09we care about most. Which leaves researchers
- 2:10basically having to study people exactly as they
- 2:13are in the wild. Precisely. We have to use correlational
- 2:17studies. But, and this is cool, we can structure
- 2:19them similarly to experiments using what's called
- 2:22a factorial design. A factorial design, okay.
- 2:25Which just means we're looking at multiple independent
- 2:27variables, but we aren't manipulating them. We
- 2:29just measure them as they naturally exist. Okay,
- 2:31let's ground this in a real example to make it
- 2:34easier to remember. Let's imagine... a hypothetical
- 2:37study where researchers want to figure out what
- 2:39drives someone's willingness to have unprotected
- 2:42sex. OK, a very real -world scenario. Right.
- 2:45So that willingness is our dependent variable.
- 2:47That's the outcome we care about. And the researchers
- 2:50want to see how two specific traits influence
- 2:52it. Let's say the person's current mood, whether
- 2:55it's positive or negative, and their baseline
- 2:58self -esteem, whether it's high or low. Got it.
- 3:00And because you have two different independent
- 3:02variables, mood and self -esteem, and each one
- 3:06has two levels, positive and negative, high and
- 3:09low researchers would conceptualize this as a
- 3:12two -by -two factorial design. Okay, so visually,
- 3:15for everyone studying this, it reminds me of
- 3:18a Punnett square from high school biology. Oh
- 3:20yeah, that's a perfect way to picture it. Right,
- 3:22you have a grid. Mood is on the left side, self
- 3:25-esteem is across the top, and you end up with
- 3:28four little boxes intersecting in the middle.
- 3:30That's a fantastic way to visualize it. And that
- 3:33grid allows researchers to look at the main effects,
- 3:36like does mood alone change the behavior, or
- 3:38does self -esteem alone change it? But more importantly,
- 3:41you look at the interaction inside those boxes.
- 3:44Does a positive mood combined with low self -esteem
- 3:46create a uniquely high risk for the behavior?
- 3:49And the psychological mechanism there totally
- 3:51makes sense, right? Like, a positive mood might
- 3:54increase someone's drive for sensation and reward.
- 3:57Exactly. And if they simultaneously have low
- 4:00self -esteem, maybe they struggle to set healthy
- 4:02boundaries in that specific moment. Right. You're
- 4:04exploring the why behind the intersection of
- 4:07those traits. Now, just for some context, there
- 4:09actually was a famous study by McDonald and Martino
- 4:12in 2002 that looked at these exact variables.
- 4:15Oh, really? Yeah, but... Their study actually
- 4:18was an experiment. They deliberately manipulated
- 4:21their participants' moods by forcing them to
- 4:24watch different videos in a lab setting. Okay,
- 4:26but in our hypothetical 2x2 scenario, we aren't
- 4:29forcing anyone to watch anything. We're just
- 4:30measuring them as they walk through the door.
- 4:32Right, purely correlational. But here is the
- 4:34danger zone, and where I kind of have to push
- 4:36back a little. Okay, go for it. If our 2x2 grid
- 4:39mathematically looks identical to a controlled
- 4:42experiment, how do we stop ourselves from jumping
- 4:45to the conclusion that, say, low self -esteem
- 4:49causes the risky behavior? Ah, that is the trap.
- 4:52You have to constantly remind yourself of two
- 4:54massive pitfalls, the directionality problem
- 4:57and the third variable problem. Even if the data
- 5:00clearly shows that people in a positive mood
- 5:03are highly correlated with taking that risk,
- 5:06we absolutely cannot infer causality. Because
- 5:08something else might be pulling the strings behind
- 5:10the curtain. Exactly. There might be a completely
- 5:13unmeasured third variable. Take impulsivity,
- 5:15for example. A highly impulsive personality might
- 5:19cause someone to experience intense, euphoric,
- 5:22positive moods. And that exact same impulsivity
- 5:26is what causes them to engage in risky behavior.
- 5:29Oh, wow. So the mood isn't causing the behavior
- 5:31at all. Right. The impulsivity is causing both
- 5:34of them. Or even something as simple as, like,
- 5:36alcohol consumption on a Friday night. It boosts
- 5:39the mood, lowers the inhibitions, and drives
- 5:41the risky choices all at the exact same time.
- 5:44Spot on. So the 2x2 grid is a really powerful
- 5:47starting point, but it requires a profound amount
- 5:52of scientific humility. It also seems just a
- 5:54bit limited. I mean, human behavior doesn't really
- 5:56happen in a neat little four -square grid. Oh,
- 5:58definitely not. Like when I make a decision,
- 6:00there are dozens of factors influencing me all
- 6:02at once. So how do researchers handle a massive
- 6:06pile of real -world variables simultaneously?
- 6:10Well, they do it by abandoning the simple grid
- 6:12entirely. They move into assessing relationships
- 6:15among multiple variables all at once. Okay, so
- 6:19it gets messy. Very. Most complex correlational
- 6:22research measures a whole bundle of variables.
- 6:25Some are quantitative, like age or income, and
- 6:28some are categorical, like gender or occupation.
- 6:31And then it assesses the statistical relationships
- 6:33among all of them concurrently. OK, that makes
- 6:36me think of the Radcliffe and Klein study on
- 6:38optimism from 2002. That's a perfect showcase
- 6:41for this, right? It really is. They were curious
- 6:44about how optimism practically affects our physical
- 6:46well -being. So they used the life orientation
- 6:49test, which is this widely accepted questionnaire,
- 6:52to measure optimism in middle -aged adults. But
- 6:54they didn't just measure optimism in, like, one
- 6:57other thing. No, no. They measured a huge spectrum
- 6:59of variables related to heart attack risks. And
- 7:02when you look at the psychological mechanics
- 7:04of what they found, it's pretty fascinating.
- 7:06The highly optimistic participants were significantly
- 7:09healthier. Like, they had lower blood pressure
- 7:11and they exercised way more often. Right, which
- 7:14sounds great, but the interesting part is why.
- 7:16Exactly. It's not magic. The studies show that
- 7:20optimistic people actively knew more about heart
- 7:22attack risk factors. They genuinely believed
- 7:26their personal risk was lower than their peers,
- 7:29which, you know, you'd think might make them
- 7:31complacent. Right. You'd think they'd just relax.
- 7:33Yeah. But it actually seemed to motivate them
- 7:35to protect their health even more. Because they
- 7:38believe their actions would actually yield positive
- 7:41results. So they put in the real effort. Yeah.
- 7:44And what's crucial about this approach you know,
- 7:47measuring dozens of variables at once, is that
- 7:49it's often how psychologists prove that a new
- 7:52test actually works. Wait, how do you mean? Like,
- 7:54if you invent a brand new psychological scale,
- 7:57you have to prove to the scientific community
- 7:59that it actually measures what you claim it measures.
- 8:02Ah, which brings us perfectly to Cassioppo and
- 8:05Petty in 1982. Yes, a classic. They wanted to
- 8:08measure something really abstract. basically
- 8:10how much people genuinely enjoy thinking. They
- 8:12called it the need for cognition. That's a great
- 8:15term. Yeah. Like some people just love solving
- 8:17puzzles and debating philosophy while others
- 8:20find that kind of mental effort completely exhausting.
- 8:23So to prove their new need for cognition scale
- 8:25was actually valid, they gave it to a massive
- 8:28sample of college students. But they also measured
- 8:31three other established variables. Right. Intelligence
- 8:33was one. Yep. Intelligence. Also socially desirable
- 8:36responding, which is basically how much a person
- 8:39lies, just to give the polite expected answer.
- 8:42Right, right. And dogmatism, which is the tendency
- 8:45to lay down principles as undeniably true without
- 8:48ever considering the evidence. Dogmatism is so
- 8:51fascinating. It's that really rigid black and
- 8:53white, my way or the highway kind of thinking.
- 8:55So to see how all these things interacted, the
- 8:58researchers created a correlation matrix using
- 9:01Pearson's R. Right. And a correlation matrix
- 9:03simply calculates the correlation between every
- 9:06possible pair of variables in the entire OK,
- 9:08so I always visualize a correlation matrix kind
- 9:11of like the mileage chart in the back of an old
- 9:13road atlas. Oh, that's a good analogy. Yeah,
- 9:15like the variables are the cities. If you want
- 9:18to know the distance between Chicago and New
- 9:20York, you just slide your fingers down the row
- 9:22and across the column until they intersect. The
- 9:25number in that specific box is your correlation
- 9:27between those two traits. There's a highly effective
- 9:30mental image. And just like a mileage chart has
- 9:34a line of zeros diagonally down the middle because,
- 9:37you know, the distance from Chicago to Chicago
- 9:39is zero. A correlation matrix has a diagonal
- 9:42line of perfect positive correlations. It's usually
- 9:46written as 1 .00 because a variable obviously
- 9:49correlates perfectly with itself. So when we
- 9:51look at the mileage between these psychological
- 9:53traits, Cacioppo and Petty's data really tells
- 9:56a story. The need for cognition correlated with
- 9:59intelligence at a positive 0 .39. Right. So the
- 10:02higher the need for cognition, the higher the
- 10:03intelligence tended to be. Exactly. But the real
- 10:05kicker was the dogmatism. It correlated with
- 10:08the need for cognition at a negative .27. Meaning
- 10:11as the enjoyment of deep thinking went up, rigid
- 10:14dogmatic beliefs went down. Which makes perfect
- 10:18psychological sense. People who truly love to
- 10:21turn an idea over in their mind, who enjoy the
- 10:24friction of complex thought, they're naturally
- 10:27going to resist dogmatic rules that tell them
- 10:29exactly what they have to think. Exactly. And
- 10:33that overall pattern of correlations, the way
- 10:35the new scale interacted with those established
- 10:37human behaviors, proved to the researchers that
- 10:39their tool was accurately capturing that underlying
- 10:42psychological trait. But let's be honest for
- 10:45a second. Looking at a matrix with four variables
- 10:47is fine. That's a tiny little grid. But what
- 10:50if a researcher measures 50 different variables?
- 10:53Right. It's messy. Right. Isn't staring at a
- 10:55massive grid of hundreds of intersecting numbers
- 10:57just pure information overload? Like, how does
- 10:59any human being find a meaningful pattern in
- 11:02that kind of noise? You're touching on one of
- 11:04the biggest challenges in data science right
- 11:06there. When the matrix gets too big, the human
- 11:09brain simply can't process it. It's impossible.
- 11:12We need a way to compress the data. We need to
- 11:15find the hidden gravity pulling certain variables
- 11:17together. And that leads us to a really brilliant
- 11:20statistical solution, which is factor analysis.
- 11:23Okay, factor analysis sounds incredibly intimidating,
- 11:26but the concept is actually pretty elegant. It
- 11:29really is. In essence, factor analysis is an
- 11:32advanced statistical technique that takes massive
- 11:34amounts of variables and mathematically organizes
- 11:37them into a smaller number of clusters. OK. The
- 11:40algorithm basically looks for patterns where
- 11:43variables are strongly correlated within their
- 11:45specific cluster, but very weakly correlated
- 11:48between the different clusters. So just as an
- 11:50example, if we measure 50 different mental tasks,
- 11:55math puzzles, reading comprehension, spatial
- 11:57reasoning, vocabulary, tests, the math doesn't
- 11:59just leave them as 50 random data points. Right.
- 12:02It looks for relationships. It notices that the
- 12:05people who score high in vocabulary also tend
- 12:07to score really high on reading comprehension,
- 12:10so it groups them together into a hidden cluster,
- 12:12a factor, which we might eventually label as
- 12:15verbal intelligence. Precisely. Each cluster
- 12:18represents an underlying psychological construct.
- 12:21And one of the most entertaining applications
- 12:23of this was the rent -for -own Gosling study
- 12:25from 2008. the one on music preferences. Oh,
- 12:28this study is so brilliant. They surveyed over
- 12:311 ,700 university students and just asked them
- 12:34to rate how much they liked 14 different popular
- 12:37music genres. They took all those thousands of
- 12:39ratings, dumped them into a factor analysis,
- 12:42and basically asked them math. to find the hidden
- 12:45patterns. And the mass naturally grouped those
- 12:4714 seemingly distinct genres into four core factors.
- 12:51Which is wild. The researchers then had to look
- 12:55at what clumped together and figure out what
- 12:57to name the factors. So they found a reflective
- 12:59and complex factor, which pulled in blues, jazz,
- 13:03classical and folk music. They found an intense
- 13:06and rebellious factor, which was rock, alternative
- 13:09and heavy metal. And an energetic and rhythmic
- 13:12factor for rap, hip hop, soul, and electronica.
- 13:15And the fourth factor is where it gets particularly
- 13:17interesting from a cultural and even a moral
- 13:19perspective. They called it upbeat and conventional.
- 13:22This factor mathematically clumped together pop
- 13:24music, country music, soundtracks, and notably
- 13:27religious music. That is such a fascinating psychological
- 13:30link, because you wouldn't immediately put, like,
- 13:33a top 40 pop anthem in the exact same category
- 13:36as a church hymn. No, not intuitively. But mathematically,
- 13:39the exact same type of person is drawn to both.
- 13:42It really highlights the psychological function
- 13:44of that music. It's about seeking comfort, community
- 13:48cohesion, and a very predictable, uplifting emotional
- 13:51arc. Exactly. It operates almost identically
- 13:54to a modern music -streaming owl group. Oh, right.
- 13:57You know when Spotify or Apple Music notices
- 14:00you listen to three seemingly different indie
- 14:02bands and suddenly it realizes they all belong
- 14:05to a hidden genre of your personality? Yes. It's
- 14:08literally wanting massive factor analyses on
- 14:10millions of users to find those hidden clusters
- 14:13of human behavior. Yep. And we see the exact
- 14:16same architecture in personality psychology,
- 14:18the famous big five personality traits. They
- 14:20weren't just invented out of thin air. They were
- 14:23discovered using factor analysis. Researchers
- 14:26measured hundreds of specific tiny human behaviors,
- 14:29things like warmth, gregariousness, activity
- 14:32level, you know, how often someone smiles at
- 14:34strangers, and the math revealed that all those
- 14:38tiny traits naturally clustered together into
- 14:41a single massive factor that we now call extraversion.
- 14:44But wait, let me ask you this. If I love classical
- 14:47music, does this mean the math permanently puts
- 14:50me in a box? Like, if my personality is drawn
- 14:53to the reflective and complex factor, am I statistically
- 14:56forbidden from liking the intense and rebellious
- 14:58heavy metal factor? Okay, that is a crucial misconception
- 15:01to clear up. There are two major caveats to factor
- 15:04analysis. Okay, what's the first one? First...
- 15:06Factors are not rigid categories. You aren't
- 15:09placed in a silo. Factors are underlying constructs
- 15:12that operate completely independently of one
- 15:14another. So it's not a seesaw. Like one goes
- 15:16up, the other has to go down. Exactly. You can
- 15:19score highly on multiple factors simultaneously.
- 15:21You can be deeply reflective and love classical
- 15:24music while also having a high threshold for
- 15:27intensity and blasting heavy metal in your car.
- 15:30Just like in the Big Five, you could be highly
- 15:33extroverted and highly conscientious at the exact
- 15:36same time. Okay, that makes sense. What's the
- 15:38second caveat? The math is entirely blind. The
- 15:41computer has absolutely no idea what jazz or
- 15:44extroversion or religious music actually is.
- 15:47It only reveals the structural shadow of the
- 15:50data. Oh, I see. It is entirely up to the human
- 15:52researcher to look at that cluster, interpret
- 15:55the psychology behind it, give it a label, and
- 15:57explain why that structure exists. Like figuring
- 16:00out what object is actually casting the shadow.
- 16:02Yes, exactly. For example, Plowman and his colleagues
- 16:05in 2008 demonstrated that the reason the big
- 16:07five traits operate as completely separate independent
- 16:10factors is likely because they're controlled
- 16:12by entirely different sets of genes. Wow. The
- 16:15factor analysis just showed us the blueprint.
- 16:16The researchers actually had to uncover the genetic
- 16:18foundation. OK, so finding hitting patterns and
- 16:21sorting messy data into neat clusters is incredibly
- 16:25powerful. But at the end of the day, psychology
- 16:27wants to solve problems. We want to know what
- 16:30causes what. So if we can't do an experiment,
- 16:33how can a correlational study ever actually tease
- 16:36out true causality? Well, the hard truth is pure
- 16:40correlational research can never definitively
- 16:42prove causality. That old mantra correlation
- 16:45does not imply causation is permanent. However,
- 16:48and this is key, we can use complex statistics
- 16:51to aggressively rule out alternative interpretations.
- 16:55We do this through the statistical control of
- 16:57potential third variables. So basically instead
- 17:00of holding the environment constant in a sterile
- 17:02lab, we measure all the messy variables in the
- 17:04real world and then we just use math to filter
- 17:06them out of the final equation. Exactly. Let's
- 17:08look at the 2011 study by Piff and his colleagues.
- 17:11It was on socioeconomic status or SES and generosity.
- 17:14Oh, I know this one. Yeah, they had a fascinating
- 17:16hypothesis. They thought that being lower in
- 17:19SES, actually having less wealth, causes people
- 17:22to be more generous. To test this, they used
- 17:24the dictator game, which is a classic psychological
- 17:27setup. It's brilliant in its simplicity. They
- 17:30bring a participant in and tell them they are
- 17:32paired with a totally anonymous partner in another
- 17:35room. But in reality, There is no partner. Right.
- 17:38Total setup. They give the participant 10 points,
- 17:41which translate to real money, and say, you are
- 17:43the dictator. You can split this wealth with
- 17:46your unseen partner however you want. You can
- 17:48even keep all 10 points for yourself, and no
- 17:50one will ever know. It's a great test because
- 17:53it strips away social pressure and isolates pure
- 17:56generosity. And the findings were really striking.
- 17:59Participants who were lower in SES reliably gave
- 18:03away more of their points than those higher in
- 18:05SES. And psychologically, you can see why. People
- 18:08with fewer resources often rely more heavily
- 18:10on community support and mutual aid, which builds
- 18:13a stronger baseline of empathy for a stranger
- 18:16in need. It aligns perfectly with their hypothesis.
- 18:19But if you are a rigorous scientist, you have
- 18:21to look for the third variable. OK, let's play
- 18:23devil's advocate then. Let's do it. If I'm a
- 18:25skeptic, I'd argue that maybe their bank account
- 18:27has absolutely nothing to do with it. Okay. Maybe
- 18:30lower SES individuals in this specific geographic
- 18:33sample just happen to be highly religious. Maybe
- 18:36it is their religious morality. you know, their
- 18:39internal sense of divine duty to the poor that
- 18:42is driving their generosity in the dictator game,
- 18:45not their socioeconomic status. That's a great
- 18:48alternate theory. Or perhaps it's tied to cultural
- 18:50norms within specific ethnic communities that
- 18:52heavily emphasize shared resources. And that
- 18:55skepticism is the engine of good science. Piff's
- 18:59team actually anticipated those exact arguments.
- 19:03That is why they didn't just measure wealth,
- 19:05they actively measured their participants' religiosity
- 19:07and their ethnicity. Right, to see if those factors
- 19:10were the true puppet masters. Exactly. And when
- 19:12they included those variables in their statistical
- 19:14analysis, they found that neither religiosity
- 19:17nor ethnicity was correlated with generosity
- 19:20in the context of this specific game. So the
- 19:23math effectively washed them out. It proved that
- 19:26in this scenario, divine duty or religious morality
- 19:28wasn't actually driving the behavior. Yes. They
- 19:32statistically controlled for those third variables,
- 19:35ruling them out as the primary cause. Now, does
- 19:38this prove with 100 % certainty that lack of
- 19:42wealth causes generosity? No. No, because there
- 19:45could always be an unmeasured fourth or fifth
- 19:47variable. But by systematically neutralizing
- 19:50the most logical alternative explanations, they
- 19:53built a highly defensible case for SES being
- 19:56the true driver. OK, but when you say they statistically
- 19:59controlled for them or included them in the analysis,
- 20:02what is the actual mechanism? Like, practically,
- 20:04how do you mathematically wash out a variable?
- 20:06Ah, this brings us to the ultimate recipe of
- 20:08complex correlational research. Multiple regression.
- 20:12Multiple regression. Yes. It's a statistical
- 20:14tool where you take several independent variables,
- 20:16the things you think are causing the behavior,
- 20:18and you measure their impact on a single dependent
- 20:20variable, which is your outcome. So in statistical
- 20:23software, it literally builds an equation. Yes.
- 20:26An additive equation. You have your outcome,
- 20:29y. And y is equal to a combination of all your
- 20:32independent variables, your x's. But here is
- 20:35the secret. Each x gets multiplied by a specific
- 20:39regression weight, which is denoted by the letter
- 20:42b. Okay, so it's not just throwing everything
- 20:44into a pot randomly. The B weight tell you exactly
- 20:46how much influence that specific variable has.
- 20:49Precisely. A regression weight indicates exactly
- 20:52how much the dependent variable changes for every
- 20:55one unit change in that specific independent
- 20:57variable while holding all the other variables
- 20:59completely constant. A really great way to conceptualize
- 21:02this is to think of multiple regression like
- 21:04baking a cake. Oh, I like this. Yeah, so the
- 21:06dependent variable, your outcome Y, is the final
- 21:09cake. It's the taste you are trying to predict.
- 21:11independent variables your X's are your ingredients
- 21:14like flour, sugar, butter. And the regression
- 21:17weights. The weights are the exact measurements
- 21:19on your measuring cups. The multiple regression
- 21:22equation tells you mathematically how much the
- 21:24sugar is contributing to the final sweetness
- 21:26of the cake, independently of the butter or the
- 21:29flour. It isolates the unique contribution of
- 21:32every single ingredient. That analogy hits the
- 21:35nail on the head. Multiple regression reveals
- 21:37if an ingredient makes a unique contribution
- 21:40over and above the contributions made by everything
- 21:42else in the bowl. Let's apply that to a really
- 21:44complex human question. The Deaner study from
- 21:472000 tried to predict human happiness. So happiness
- 21:50is our cake. He used two independent variables,
- 21:53income and health. The problem is, wealth and
- 21:57health are deeply tangled together. Very tangled.
- 22:00Richer people can afford better health care and
- 22:01gym memberships. Healthier people can work longer
- 22:04hours and earn more wealth. So if a wealthy person
- 22:07reports being very happy, Is it the money making
- 22:10them happy or is it the fact that they are healthy
- 22:12enough to enjoy life? Multiple regression untangles
- 22:15them. It calculates whether income makes a unique
- 22:18contribution to happiness when health is mathematically
- 22:21frozen and vice versa. And the results of that
- 22:25study are incredibly humbling. The regression
- 22:27weights showed that both income and health actually
- 22:31make extremely small, unique contributions to
- 22:34your overall happiness. Surprising, isn't it?
- 22:36It is. The only time they become major drivers
- 22:38of happiness is if you are at the extreme ends
- 22:41of the spectrum, like living in severe poverty
- 22:43or battling a severe, debilitating illness. It
- 22:46teased apart two deeply intertwined variables
- 22:49to find the unique psychological weight of each.
- 22:52But we really have to end on the golden rule
- 22:55of this entire field. Which is? Even with the
- 22:57perfect multiple regression recipe, purely correlational
- 23:01research still cannot unambiguously establish
- 23:04causation. Right. It just shows patterns of relationships
- 23:06that are highly consistent with a causal interpretation
- 23:09and inconsistent with others. Exactly. So let's
- 23:11step back and look at what we've decoded today
- 23:13in this study session. We started with the constraints
- 23:16of the real world, learning how to use two by
- 23:19two factorial grids to observe behavior without
- 23:22manipulating it. We mapped the distance between
- 23:24psychological traits using correlation matrices,
- 23:27finding that deep thinkers often resist dogmatism.
- 23:31We learned how factor analysis finds the hitting
- 23:33gravity pulling our behaviors into traits, even
- 23:36revealing why pop music and religious hymns share
- 23:39the same psychological architecture. It's all
- 23:41connected. And finally, we explored how multiple
- 23:44regression allows us to mathematically isolate
- 23:47variables, proving whether it's our bank account
- 23:50or our religious morality driving our gen...
- 23:52generosity. For you listening, you are walking
- 23:55away with the analytical tools to be a true critical
- 23:58thinker. The next time you see a headline making
- 24:01a bold claim about human nature, you don't have
- 24:04to just blindly accept it. Definitely not. You
- 24:07should immediately ask yourself, did they run
- 24:08a multiple regression? Did they mathematically
- 24:10control for third variables? What else could
- 24:13be causing this? Exactly. Don't just read the
- 24:15headline. Look for the regression weights. But
- 24:18I want to leave you with one final kind of provocative
- 24:21thought today. All right. Lay it on us. We talked
- 24:24a lot about the hidden math of behavior. If everything
- 24:27from our deepest moral choices of sharing wealth
- 24:29with a stranger to the spantual comfort we seek
- 24:32in religious music, if all of that can be cleanly
- 24:35mapped out into mathematical clusters and predictable
- 24:37regression equations, at what point does human
- 24:41free will just become a highly complex but entirely
- 24:44predictable statistical formula? That is a heavy
- 24:47question to end on. It really is. Keep questioning
- 24:50the data, keep questioning the world around you,
- 24:51and we'll see you next time. on the Psychology
- 24:53undergrad.