Week 5 Reading Post — The Truthful Art: Chapter 8 and Chapter 9
Chapter 8: Revealing Change
Cairo begins this chapter with an example about Spain’s unemployment rate and an infographic that showed unemployment rates decreasing between February and August of 2013. Critics and Cairo point out that Spain’s jobs are highly seasonal, with jobs increasing in the winter months, and that charts comparing unemployment rates between August 2012 and August 2013, the unemployment actually increases. This shows the importance of displaying information in time series line charts while examining things like trend, seasonality, and noise in the variations. These are especially important to consider when looking at long-term trends in data. Cairo continues with examples about Spain’s Social Security affiliates, also used to analyze country workforce, population size and distribution, and sales, focusing on seasonality fluxes and deviations from the moving average (the average based on time increments). Seasonal subseries charts are especially effective for visualizing yearly data like climate science data, company sales, and other long-term data distributions. Cairo then moves on to discuss index origins and uses single-family house sales as an example (Figure 8.16). I find these visualizations quite informative and useful because they show each line highlighted in red against the other lines in gray. I think this is a particularly helpful method in visualizing how trends compare to one another on a larger scale. Cairo also explains the logarithmic scale, something I’d forgotten about since 12th grade math class, and how it is used to show rates of change in a more constant fashion, using increments that represent magnitudes of increase (i.e. 1 increment in a log10 scale represents a tenfold increase). Cairo finishes the chapter by highlighting the importance of avoiding paradoxes and displays some very cool infographic chart types at the end such as horizon charts, connected scatter plots, and more time series line charts.
Chapter 9: Seeing Relationships
Cairo begins this chapter with the humorous example of Spurious Correlations, which I research a little after reading this — it’s pretty entertaining, and variables that appear to be associated in absurd ways. Many of the variables given in these examples are related to a country’s wealth, so while association does appear to occur, it isn’t due to the variables affecting one another. Association occurs when a change in one variable is accompanied by changes in the other variable, and if these points are plotted along a trend line: the closer they are to the trend line, the stronger to association between variables. Correlation can occur when the relationship between two variables is linear, or when it is exponential. The correlation coefficient (-1.0 < r < 1.0) is an important value to calculate when examining the relationship between two variables. Cairo then moves on to discuss locally weighted scatter plot smoothing, using the example of SAT scores that are distributed unevenly because not everyone in each region of the country takes the SAT. Another way to visualize multivariate data, data sets with data points for multiple variables, is the parallel coordinates plot. Cairo points out an important discrepancy in the titles of an example on the Bolsa Família welfare program and voting for Rouseff. The titles imply a correlation, and while a relationship does exist, it does not point directly to causation. I think this is a really important factor to consider when examining any data set. I think it is also easy to convince yourself that correlation and causation exist, simply because, as Cairo pointed out in earlier chapters, our brains like to make connections. For this reason, I think it’s important to take a step back and try to look at the data more objectively or bring in alternate viewers to provide additional opinions. Cairo ends this chapter by explaining different variables in the regression model, specifically going into depth about the univariate linear least squares regression, and discussing more about the relationship between correlation and causation.
