Psychology for Writers: Amnesia

When I started getting ready for this blog series, I added some posts to writer groups I belong to, asking people to describe some of the tropes in fiction that drive them crazy. One person commented on how head injuries are portrayed in fiction, and many replied with similar frustrations: people are knocked unconscious for a time and wake up sometime later with no lingering issues, bumps to the head are used to recover memories, and of course, the whole "don't let someone suffering from a concussion fall asleep" nonsense.

Two jobs ago, I worked as a researcher for the Department of Veterans Affairs. Though I didn't study this topic of head injuries myself, many of my colleagues did. (I've also suffered a couple of those in my lifetime, so I can talk about some of the experience firsthand.)

Traumatic brain injury, or TBI, is the medical term for a head injury. I was originally going to write this post on TBI, but realized that would make for a very long post - or else, a post in which I gloss over a lot of detail. So instead, I'm going to talk specifically about amnesia.

TBI is usually classified into one of three categories: mild, moderate, and severe. But any insult to the brain can bring with it side effects. The person suffering the head injury will probably experience what's called "post-traumatic amnesia," where they may lose their memory of things happening just before and/or just after the event. And they probably won't remember the experience of hitting their head at all. For a mild head injury, the individual may have a more severe bout of amnesia that comes on later, but will still be short-term. In fact, the delay between the head injury and this amnesia can also be seen as an indicator of severity. For more severe head injuries, amnesia and other neurological symptoms, like seizure, may be almost immediate.

The length of this experience of amnesia is also related to severity. For extremely mild injuries, post-traumatic amnesia may last only a few minutes. For a severe injury, it can last days.

And no, another blow to the head won't bring the memory back. In fact, a new head injury shortly after an old one is very bad, because the brain can't heal itself as well. Too many injuries, and the brain may be unable to heal itself at all. This is why retired professional football players have disorders that look like dementia. Their brain has essentially become scar tissue.

For me, the amnesia happened about 3 hours after the injury - I couldn't respond to questions about who I was, who my family was, the day, or my dog's name (all questions my mom asked me). I remember this feeling odd but not as terrifying as you might think. In fiction, amnesia sufferers wander around terrified, asking who they are. But for me, it was more like the answers to these questions were on the tip of my tongue, and with just a little more time, I could answer all of them. Less "Oh my god, I don't remember!" and more "That's strange. I swore I knew that." So at some level, I knew that I had the answers to these questions somewhere in my brain. 28 years later, I still remember all of this, and all of the questions I was asked at the time. The only memory I have never recovered is the second when my head came in contact with the wall. (The experience in the ER is also a bit fuzzy, but mostly because I kept falling asleep - I was exhausted by that point. And I slept a lot in the months after, as I recovered.)

I was feeling back to normal, memory-wise, within a few hours. Now, there are longer-term side effects of a head injury - things like fatigue, insomnia, and intermittent memory issues, which comprise what's known as post-concussive syndrome. Another post for another day.

So amnesia via head injury is generally short-term unless the person is severely injured, and at that point, a little amnesia may be the least of the person's worries.

It's actually much more believable for someone to experience amnesia as a result of some kind of psychological trauma. Dissociation is a psychological term for when a person loses his or her sense of self, and encompasses a handful of related disorders. What was previously known as multiple personality disorder is a dissociative disorder, now called dissociative identity disorder. Dissociation becomes a defense mechanism to protect oneself from re-experiencing trauma.

During dissociative fugue - a much better explanation if you want a character to wander into a town with no memory - the person will often run away from their life and commitments. (In fact, the word "fugue" comes from the Latin word for flight.) While in this state, the person may forget who they are and details of their life. It isn't unusual for a person in a fugue to assume another identity, likely because they realize they need one to function. They probably won't appear to be in immediate distress. After all, the fugue was brought on to escape distress, and to most observers, the person will appear normal. The person may eventually come out of the fugue, but it's harder to predict. Unlike a head injury, which has a physical cause, and therefore a clear trajectory of healing, the block here is psychological and will exist as long as the person needs protection from whatever is causing distress.

Finally, there's a type of dissociation/amnesia many have experienced - alcohol-induced blackouts. Once again, short-term and physical in cause. The hangover will probably last longer than the amnesia itself.

Releasing an Oscar-Worthy Movie

Apparently, having an Oscar-worthy movie isn't just about the script, the cast, or the director - it's also about how the movie is released. Market Match published this great story last week, that describes the two approaches to releasing movies:

A wide release is usually pursued when a distributor believes a movie will be an immediate hit, on the basis of its existing fandom or a major marketing campaign.

The platform release, on the other hand, can be used to gauge a film’s appeal. “A platform release takes a lot of planning. The strategy usually changes on the fly through the response of public and critics,” said Paul Dergarabedian, a senior media analyst with comScore.

This year, only three of the Best Picture nominees used a wide release approach: Dunkirk, Get Out, and The Post. The rest (Three Billboards Outside Ebbing, Missouri, Darkest Hour, Lady Bird, The Shape of Water, Call Me By Your Name, and Phantom Thread), used the platform release approach.

And in a fantastic visual display that you really need to see for yourself (scroll partway down), the authors demonstrate that, over the last 8 years, 7 of the Best Picture winners used the platform release approach.

Statistical Sins: Types of Statistical Relationships

On this blog, I've covered a lot of statistical topics, and have tried to make them approachable even to people with little knowledge of statistics. But I admit that, recently, I've been covering more advanced topics. But there are still many basic topics to explore, that could be helpful for non-statistical readers, as well as teachers of statistics who address these topics in courses. This topic was prompted by a couple of discussions - one late last week, and another at dance class last night.

The discussions dealt with what it means to say two variables are related to each other (or not), including whether that relationship is weak, moderate, or strong, and whether it is positive or negative. I first addressed this topic when I wrote about correlation for my 2017 April A to Z. But let's really dive into that topic.

You may remember that correlation ranges from -1 to +1. The closer the value is to 1 (positive or negative), the stronger the relationship between the two variables. So in terms of strength, it is the absolute value that matters. A correlation close to 0 indicates no relationship between the variables.

A common question I see is whether a correlation is weak, moderate, or strong. Different people use different conventions, and there isn't a lot of agreement on this topic, but I frequently see a correlation of 0.5 referred to as strong. As I explained in my post on explained variance, this means the two variables share about 25% variance, or more specifically, about 25% of the variation in one variable can be explained by the other variable.

For moderate, I often see 0.3 (or 9% shared variance) and for weak, 0.1 (or 1% shared variance). But as I said, these really aren't established conventions (please don't @ me - thanks in advance). You could argue that there are a variety of factors that influence whether a relationship is seen as weak, moderate, or strong. For instance, study methods could have an impact. Finding a correlation of 0.5 between two variables I can directly manipulate and/or measure in an experimental setting is completely different from finding a correlation of 0.5 between two variables simply measured in a natural setting, where I have little to no control over confounds and high potential for measurement error. And we are often more generous about what we consider strong or weak in new areas of research than in well-established topics. But conventions give you a nice starting point, that you can then shift as needed depending on these other factors.

Direction of the relationship is indicated by the sign. A positive correlation means a positive relationship - as scores on one variable go up, so do scores on the other variable. For example, calories consumed per day and weight would be positively correlated.

A negative correlation means a negative relationship - as scores on one variable go up, scores on the other variable go down. For instance, minutes of exercise per day and weight would probably have a negative correlation.

[Now, for that last relationship, you might point out that there are a variety of other factors that could change that relationship. For example, some exercises burn fat, lowering weight, while others build muscle, which might increase weight. I hope to explore this topic a bit more later: how ignoring subgroups in your data can lead you to draw the wrong conclusions.]

Part of the confusion someone had in our discussion last night was knowing the difference between no relationship and a negative relationship. That is, they talked about how one variable (early success) had no bearing on another variable (future performance). They quickly pointed that this doesn't mean having early success is bad - the relationship isn't negative. But I think there is a tendency for people unfamiliar with statistics to confuse "non-predictive" with "bad".

So let's demonstrate some of these different relationships. To do that, I've generated a dataset in R, using the following code to force the dataset to have specific correlations. (You can use the code to recreate on your own.) The first thing I do is create a correlation matrix, which shows the correlations between each pairing of variables, that reflect a variety of relationship strengths and directions. Then, I impose that correlation matrix onto a randomly generated dataset. Basically R generates data that produces a correlation matrix very similar to the one I defined.

R1 = c(1,0.6,-0.5,0.31,0.11)
R2 = c(0.6,1,-0.39,-0.25,0.05)
R3 = c(-0.5,-0.39,1,-0.001,-0.09)
R4 = c(0.31,-0.25,-0.001,1,0.01)
R5 = c(0.11,0.05,-0.09,0.01,1)

R = cbind(R1,R2,R3,R4,R5)
U = t(chol(R))
nvars = dim(U)[1]
numobs = 1000
set.seed(36)
random.normal = matrix(rnorm(nvars*numobs,0,1),nrow = nvars, ncol=numobs)
X = U %*% random.normal
newX = t(X)
raw = as.data.frame(newX)
names(raw) = c("V1","V2","V3","V4","V5")
cor(raw) 

The final command, cor, which requests a correlation matrix for the dataset, produces the following:

           V1          V2         V3          V4          V5
V1  1.0000000  0.57311834 -0.4629099  0.31939003  0.10371136
V2  0.5731183  1.00000000 -0.3474012 -0.26425660  0.04838563
V3 -0.4629099 -0.34740123  1.0000000  0.01204920 -0.12017036
V4  0.3193900 -0.26425660  0.0120492  1.00000000  0.01202121
V5  0.1037114  0.04838563 -0.1201704  0.01202121  1.00000000

Just looking down the columns, you can see that the correlations are very close to what I specified. The one exception is the correlation between V3 and V4 - I asked for -0.001 and instead have 0.012. Probably, R couldn't figure out how to generate data with that correlation while also coming close to the other values I specified. So it had to fudge this one a bit.

So now, I can plot these different relationships in scatterplots, to let you see what weak, moderate, and strong relationships look like, and how direction changes the appearance. Let's start by looking at our weak correlations, positive and negative. (Code below, including lines to add a title and center it.)

library(ggplot2)

weakpos<-ggplot(raw,aes(x=V1,y=V5)+geom_point()
+labs(title="Weak Positive")
+theme(plot.title=element_text(hjust=0.5))

weakneg<-ggplot(raw,aes(x=V3,y=V5)+geom_point()
+labs(title="Weak Negative")
+theme(plot.title=element_text(hjust=0.5))

That code produces these two plots:

With similar code (just switching out variables in the x= and y= part), I can produce my moderate plots:

and my strong plots:

As you can see, even a strong relationship looks a bit like a cloud of dots, but you can see trends that go from almost nonexistent to more clearly positive or negative. You can make the trends a bit easier to spot by adding a trendline. For example:

strongpos+geom_smooth(method=lm)

The flatter the line, the weaker relationship. Two variables that are unrelated to each other (such as V4 and V5) will have a horizontal line through the scatter:

I'll come back to this topic Sunday (and in the future post idea I mentioned above), so stay tuned!

Interesting Analysis on Reasons for Congressional Resignations - And the Impact of #MeToo

According an interesting article from FiveThirtyEight, more people are resigning from Congress recently than we've seen at any other point in the last 117 years:

The 115th Congress owes its historic turnover to the confluence of two events, one normal and one abnormal. First, there’s the start of a new presidential administration. Five of the first six members to resign this session did so to accept jobs in President Trump’s administration. That’s not unusual. It’s similar to the seven members who resigned in 2009 to join the Obama administration and the five members who left in 1993 to join Bill Clinton’s.

But in addition, three of the four most recent members to resign from the 115th Congress did so because they were accused of unwanted sexual advances: John Conyers, Trent Franks and Al Franken.

The article features this graph, which shows number of resignations and reasons over time, annotated with major events occurring at that time:

Statistics Sunday: Quantile Regression (Part 1)

About a month ago, I introduced the concept of quantiles - ways of dividing up the distribution of scores into certain percentile groups. The median is one of the most well-known quantiles - it is the score that divides the distribution of scores in half, so that half of the scores in the distribution are below it and half are above. Unlike mean and standard deviation, which are most useful when scores fall along a normal distribution, quantiles can be used to describe just about any distribution (although visualization of your data distribution is still very important).

Regression is a useful technique when you want to predict an outcome based on one or more variables thought to influence that outcome. The most well-known type of regression is linear regression, where the relationship between two variables can be plotted on scatterplot with a straight line. So a linear relationship is one important assumption when using linear regression. You wouldn't want to use it if you believe the relationship between your two variables follows a curve.

But linear regression has other assumptions, one of which is that the variance of scores is consistent across the line. For instance, take a look at this scatterplot with a prediction line drawn through it:

This is the scatterplot I used in my scatterplot post, which was created using data from my study of Facebook use. Most of the dots don't fall directly on the line. And that's to be expected, because even strong predictors can't predict an outcome perfectly. When we say the relationship is linear, we are referring to the trend. But scores are going to vary and differ from our predicted score - sometimes because of other variables we didn't measure, and sometimes because of idiosyncrasies in the data. But to use linear regression, we have to meet an assumption called "homoscedasticity." That means that the variance of scores at one point on the line should be equal to the distribution of scores at other points along the line. You would want the scatter of the scatterplot to be fairly consistent across the chart. If you have a plot that looks more like this:

you're likely violating an assumption. The scores are more spread out in the upper-right corner than they are in the lower-left corner.

When I was a psychometrician at Houghton Mifflin Harcourt, we regularly worked with data that had a curved relationship between variables. That is, rather than following a straight line, the best way to describe the relationship between two variables was with a curved line - specifically, a growth curve, where we see rapid increases between two correlated variables, followed by a flattening out. There is a type of regression - polynomial regression - that allows us to model curvilinear relationships. But there was another interesting observation in these growth curves: the variance in scores at the low end was much greater than at the high end. Standard regression wouldn't do. We needed a regression approach that could handle heteroscedasticity.

Enter the quantile regression. This approach has no assumptions on distribution of scores across the regression line, because it actually gives you multiple regression lines - one for each quantile. That is, the outcome variable is divided up into however many quantiles are necessary. You could use quartiles for instance: the scores that divide the distribution into evenly sized quarters. The analysis then produces one regression equation for each division of the scores, which provides the scores that fall within each quantile. This approach is useful for situations in which the relationship between variables is not quite linear and/or when the variance is not quite consistent.

Next week, I'll provide a real example of quantile regression, and show how to conduct this analysis in R, using the quantreg package. Stay tuned!

Psychology for Writers: On Fight, Flight, and Fear Responses

This will be a new series on my blog, coming out on Fridays. (I may not post one every single week, but they'll always come out on a Friday.) My goal is to impart some wisdom from my discipline - psychology - to help writers, in part because I see a lot of tropes over and over that are not really in line with what we know about human behavior. (And I have so many writer friends with their own expertise - I'd love to see others share some domain knowledge like this to help us write more accurately!)

One of the big tropes I see is how people respond to extreme fear, usually by soiling themselves in some way. George R.R. Martin just loves to have his characters pee and crap themselves, but he's certainly not the only offender. In fact, I just finished Sleeping Beauties by Stephen King (one of my favorite authors) and Owen King, and was really frustrated when once again, the peeing-oneself-in-fear trope reared its ugly head.

Do people pee themselves in fear? Yes they do. It probably doesn't happen anywhere near as often as it does in books, but then we don't usually encounter some of literature's most fear-inducing creatures in real life. So we'll say for the moment that the frequency with which people pee themselves is fine. In fact, that's not even the issue I have.

The issue I have is when this response occurs. In the books (and movies & TV series too), something scary shows up and the character immediately wets him or herself. Nope. That's not when it happens. It would happen after the fear moment has passed - after the creature has been beaten or moved onto another target, just as the character is beginning to breath a sigh of relief. That's right - you pee after you've been afraid, not during.

How could that be? Let me talk to you for a moment about the fight vs. flight response, for which I've created this visual aid:

Regardless of whether you're going into battle or running way, your body response is the same: in response to some stressor, your body is raring up to go. Specifically, the sympathetic nervous system is getting you ready to expend energy, because fighting or flighting, you're going to be expending energy. Your body releases chemicals, like epinephrine (adrenaline), norepinephrine, and cortisol. Your heart rate and blood pressure increase. Your pupils dilate. Your muscles get ready for action.

But as these systems are charging up, other systems are slowing or shutting down completely. Your immune response is suppressed. Sexual arousal is suppressed. Digestion slows or may even stop. And your body stops producing waste materials, like urine. It's not very helpful if, just as you're getting ready to fight for your life, you realize you're hungry or need to pee, or if your allergies act up and you've suddenly got itchy eyes, sneezes, and a runny nose. Those feelings go away, temporarily, so you can focus your energy on the surviving.

But after the stressful period ends, your parasympathetic nervous system takes over. Unlike the sympathetic nervous system's "fight or flight" response, the parasympathetic nervous system's response is "rest and digest." (I've also heard "feed and breed," because of the aforementioned impact on sexual arousal.) Those processes that were slowed or stopped start up again. Your muscles relax. You can imagine where this is going.

The best portrayal of this concept I've ever seen was in The Green Mile: the scene where Wild Bill, one of the inmates on death row, grabs Percy, one of the guards, and threatens him. The other guards get Wild Bill to release Percy, who rushes to the other side of the corridor and collapses on the floor. Wild Bill says with a smirk that he wouldn't have hurt Percy. And that's when Del, a fellow inmate, points out that Percy has just wet himself.

Percy is away from Wild Bill and safe. While the emotions he was experiencing when Wild Bill grabs him are still present (he's still upset), his body is returning to its previous physiological state. And that's when his bladder lets go.

So there you have it - if you really want your character to wet him- or herself, it should happen after the character begins "coming down."

I know one question I've gotten when discussing this in the past is, why do animals pee on someone/something in fear? Remember, animals don't have the same highly evolved nervous system that we have. Their systems are not going to shut down in the same way a human's does. Also, their reaction is not directly to fear, but is (arguably) an evolved response that is pretty effective in getting predators to let them go. Sure, an animal could bite instead, but that's aggressive and likely to earn an aggressive response. These animals are generally built for flight, not fight, and peeing is a submissive response. But for humans and other creatures higher on the food chain, being an apex predator means we'll have much different responses than animals that are not apex predators.

Bonus: you could use this information to portray the opposite. Sometimes authors want to show how badass their character is by having them be completely unfazed by something another character finds terrifying. You could have your character walking into battle commenting that he/she is starving and hopes to get something to eat after finishing with this nonsense, or something along those lines.

I know I blog about statistics a lot on this blog, but I'm always thrilled to talk about my discipline too! Any psychology topics you'd like to see addressed here? (Or any tropes you're tired to seeing?)

Data Science and Trump's Tweets

A while back, I shared a post by David Robinson of Variance Explained, where he conducted a sentiment analysis on Trump's tweets. He was able to find differences in tweet content and sentiment, and showed that you could tell which tweets Trump wrote and which ones were written by someone else.

Recently, David Robinson was featured on the Data Crunch podcast, talking about his work. You can listen through the embedded audio below, or head over to this page to find out more.