topic badge

Investigation: Using technology to fit a line

Lesson

Are you in the smartest generation ever?

Research shows that the IQ of young people is increasing over time, which means that you are a member of the smartest generation ever. As a whole, your generation is more intelligent than your parents, grandparents, or any generation that ever came before it.

This steady rise in IQ scores over time is called the Flynn effect and has been the subject of a great amount of research. Many people have a deep suspicion that the Flynn effect is some sort of statistical mistake, especially at a time when many adults grumble about how "kids these days" are vastly inferior to how kids "used to be" (although, adults have been saying that for thousands of years). However, the Flynn effect is real, having been confirmed using different IQ tests in a variety of countries, from American school-children to young people in rural Kenya.

In this investigation, you're going to have a look at the figures yourself, to see what the statistics show about the rise in intelligence.

Year Average IQ
1947 100
1972 107
1989 113
2001 117

This data is adapted from Flynn's "American IQ Gains From 1932 to 2002: The WISC Subtests and Educational Progress". Basically, the way IQ tests work is that 100 is set as the average. However, when the WISC IQ test (an IQ test for school-aged people) updated their test in 1972, 1989 and 2001, they found that the average of students had crept up. Therefore, they had to "standardise" the new version of the test so that it was harder than the previous one, in order to make the average 100 again. What this table represents is the way that average IQ would have changed if these "standardisations" had not taken place.

Just from the numbers alone, you can probably see the trend: IQ scores are going up. But how is it increasing? Is it linear, exponential or somewhat random? Is it speeding up or slowing down? It is hard to tell from the numbers alone, so we are going to do some analysis to find out.

Note that we are working with a dataset that has only 4 points. We would normally need more data if we wish to produce reliable conclusions. However, for this exercise we are just exploring the concepts, which will be easier with a small dataset.

To analyse this data, we are going to use spreadsheet software such as Microsoft Excel or Google Sheets. There are a few options available so ask your teacher if you're unsure. This investigation uses and has screenshots from a specific program, but most programs function similarly. Begin by pasting the data into the spreadsheet, so that it looks like this:

Next, we're going to make a scatterplot of these figures. Click on "Insert" up the top, then "Chart". In the top of this box, you can see the tabs "Setup" and "Customise". Click on "Setup" and then click on "Scatter" from the "Chart type" pulldown menu. Here's a picture:

Now you should have a nice chart. Click on the chart title to give it a sensible name:

At first glance, the scatter plot looks like IQ is going up in pretty much a straight line. To make it even clearer, let's add a trend line. Right-click on the scatter plot, and then click "Advanced edit...". Scroll the left panel down until you get to the bottom. At the very bottom, you should see "Trend line". Click on this and select "Linear".

You should get a nice "line of best fit" on your graph. A lot easier than doing it with a ruler in Science class!

Looks like a pretty good match. And if that's all we were interested in, we could probably close the book on this one. But sometimes it's nice to be able to put a number to these things, something which allows you to easily communicate the result to other (mathematically-inclined) people. This is very important if you want to go home and tell your parents how much smarter than them you are!

For our next step, we're going to get a number, called the Pearson Correlation Coefficient, which shows how the data are related, or "correlated". The Pearson Correlation Coefficient is given the symbol r.

To understand correlation coefficients you need to know what correlation means.

The value of a correlation, the constant r, ranges between -1 and +1. A negative sign means a negative correlation, and a positive sign means a positive correlation. Can you think of an example when the correlation coefficient would be exactly +1, which means a "perfect" positive correlation?

Generally, the following terms are used to describe correlation coefficients:

Correlation Coefficient Strength
+1 Perfect positive relationship
> +0.8 Strong positive relationship
> +0.5 Moderate positive relationship
> 0 Weak positive relationship
0 No relationship
< 0 Weak negative relationship
< -0.5 Moderate negative relationship
< -0.8 Strong negative relationship
-1 Perfect negative relationship

The spreadsheet can calculate the correlation coefficient for you in the blink of an eye, using the CORREL() function. Type the following command into any empty cell: "=CORREL(A1:A5, B1:B5)" (Without the quotation marks). If your data is in a different spot to the example, you'll need to change it to reflect this. In this command, "A1:A5" means "A1 to A5", and is the location of the first set of data (in our case, year numbers). "B1:B5" means "B1 to B5" and is the location of the second set of data (in our case, IQ). If your data is in different addresses, just change these numbers so they match with the location of your data.

Once you hit enter, the correlation coefficient should come out as 0.998. In other words, a nearly perfect positive relationship. So there you have it, the relationship between Year and Average IQ is pretty much perfect. If you do the calculations, you'll see that the rate of increase is approximately 3 points of IQ every ten years. That puts you guys on an average IQ of around 120 by 1940 standards, which back then would earn you an intelligence classification of "Superior".

As well as discovering that kids are getting smarter, Flynn also discovered that they are getting smarter in some ways more than others. The IQ tests measure groups of different subcategories, focusing on different kinds of intelligence.

Here are the results from one of these subcategories, called "information". The "information" category focuses on your knowledge of memorised facts, like "What is the capital of Brazil?" or "How many planets are there in the solar system?". Here are the results of this ability over time:

Year Score
1947 100.0
1972 102.0
1989 100.5
2001 102.0

Doesn't look so impressive, does it? Copy and paste the data into the spreadsheet, and follow the same procedure to generate a scatterplot with a trend line.

The spreadsheet thinks that overall it seems to be going up, but it's definitely not as strong a relationship as what we've seen before. Use the "CORREL" command again, this time writing in the addresses of the data yourself. You should get a result of 0.59. In other words, a moderate positive correlation, but not a strong one.

So you're not really much better at remembering facts than your grandparents were. However, whilst the "Information" subcategory has improved the least out of all of the subcategories, the "Similarities" subcategory has improved the most. This category measures your ability to perform abstract reasoning by asking you questions relating to how two things are similar, for example, "How are a rabbit and a dog similar?".

Here is the data for the "similarities" subcategory:

Year Score
1947 100.0
1972 113.0
1989 119.5
2001 124.0

Do the same thing again with the spreadsheet, plotting the data and a trend-line, and calculating a correlation coefficient. What does the correlation coefficient tell you? Notice how the improvement in IQ is close to 5 IQ points per decade. At this rate, if IQ were based solely on the performance in the similarities subcategory, the average IQ for your generation would be above 130, which is classified as "very superior".

So why is it that your generation is "superior" overall, "average" at remembering stuff, and "very superior" at abstract reasoning? Well, that's the tricky part. One popular theory is that young people's environment is generally more intellectually stimulating than before, both inside and outside of the classroom. You are getting a better education than ever before at school and you are using technology like computer and the internet, which require high-level abstract reasoning to use. This intellectually stimulating environment could explain why you are not much better at fact recall, but a whole lot better at abstract reasoning.

Of course, this just one theory, and no-one really knows whether this is the right answer. Since your generation is apparently the smartest one so far, see if you can figure it out for yourself! Have a think about the following things, in a group discussion or just by yourself:

1. What does a rising IQ mean for society?

2. How are you going to convince your parents that you are smarter than they are? (Hint: try saying "Longitudinal data from American school cohorts in the period 1947 to 2002 show a linear increase in IQ with a Pearson Correlation Coefficient of 0.998". Even if they don't understand what you mean, it sounds smart enough to prove your point)

If you want to read more about the Flynn effect, read this article, or watch James Flynn's TED Talk on the topic.

 

Another analysis - does money make people happy?

In 2003, a poll called the "Gallup World Poll" asked people from 160 different countries the following question:

“All things considered, how satisfied are you with your life as a whole these days? Use a 0 to 10 scale, where 0 is dissatisfied and 10 is satisfied". The following data shows the result in a few of these countries, as well as those countries' GDP (which roughly measures how much money its citizens make per year).

Country GDP Mean Life Satisfaction
Brazil 7204.94 6.7
China 4969.64 4.5
Denmark 27969.73 8.0
India 2990.07 5.3
Japan 24036.74 6.6
Russia 11788.58 4.9
Spain 20644.31 7.2
Togo 789.04 3.3
United Kingdom 26045.56 7.0
United States 34875.37 7.1
Venezuela 6253.34 7.2

Over to you! Plot the result on a spreadsheet, make a scatterplot, add a trend-line, get a correlation coefficient, and see if you can interpret it. More importantly, see if you can critique this approach: what does this data really say about the relationship between money and happiness? Can we trust it? What other evidence would you want to see? Should we even be analysing things like this mathematically, or should we be leaving it to the philosophers? Does the answer matter enough to require a scientific response, or should it just be a matter of personal opinion?

If you want to read more about the mathematics of the relationship between money and happiness, you can read this article (which has groups of graphs and maths), or this one (which has no graphs or maths).

Just keep in mind that correlation does not imply causation, lest you end up discovering that eating more chicken causes higher oil imports.

Outcomes

ACMEM142

use technology to find the line of best fit

ACMEM144

use technology to find the correlation coefficient (an indicator of the strength of linear association)

ACMEM146

recognise the dangers of extrapolation

What is Mathspace

About Mathspace