Feeds:
Posts

Archive for the ‘Statistics’ Category

Anecdotes, the topic of the post at this link, are often used as evidence for claims, even though they have weaknesses that limit severely their value as evidence. Anecdotes are important for another topic we’re discussing in my PSY 101 course: the confirmation bias, which is a strong tendency to readily accept evidence that seems to support beliefs we already have (i.e., our preconceptions) and to examine closely evidence that seems to contradict our preconceptions in order to find problems with the disconfirming evidence that allow us to discount it.

Figure 1. Discounting evidence that disconfirms one's preconceptions can lead to the development of bizarre beliefs that people hold fervently.

Figure 1. Discounting evidence that disconfirms one’s preconceptions can lead to the development of bizarre beliefs that people hold fervently.

In everyday life, however, we often do not even become aware of disconfirming evidence. Sometimes, this is because we can easily avoid it (e.g., we might avoid watching certain news programs that we know will make claims that contradict our beliefs). Other times, our cognitive limitations make it difficult for us to notice the disconfirming evidence (i.e., it doesn’t reach the conscious level). One example of the latter involves the belief that washing our cars causes it to rain. Many of us are able to point to occasions (anecdotes) on which it rained after we washed their cars, which seems to be compelling evidence of the truth of this belief. We often fail, however, to notice (and, therefore, don’t remember) occasions on which these events didn’t occur together. In order to get better evidence for the accuracy of the belief, we would need to make the kinds of observations indicated in the following table.

Table 1. Each cell of this table indicates the observations that would need to be made to determine if washing one's car causes it to rain.

Table 1. Each cell of this table indicates the observations that would need to be made to determine if washing one’s car causes it to rain.

In making observations that would allow us to fill in the cells of the table, we force ourselves to pay attention not only to evidence that supports our belief, but also to evidence that might disconfirm that belief:

  • The cell labelled A shows the number of times we washed the car and it rained.
  • The cell labelled B shows the number of times we washed the car and it didn’t rain.
  • The cell labelled C shows the number of times we didn’t wash the car and it rained.
  • The cell labelled D shows the number of times we didn’t wash the car and it didn’t rain.

Let’s say that we make the relevant observations for one year and get the following results.

Table 2. The cells of the table show, for a one-year period, the proportion (percentage) of days we either washed or did not wash the car and it either rained or did not rain.

Table 2. The cells of the table show, for a one-year period, the proportion (percentage) of days we either washed or did not wash the car and it either rained or did not rain.

Was it more likely to rain on the days we washed the car?

  • Cell A shows that, on 20% of the days, we washed the car and  it rained.
  • Cell B shows that, on 80% of the days, we washed the car and it didn’t rain.
  • Cell C shows that, on 20% of the days, we didn’t wash the car and  it rained.
  • Cell D shows that, on 80% of the days, we didn’t wash the car and it didn’t rain.

Thus, regardless of whether we had just washed our car or not, it rained on 20% of the days that year (and it didn’t rain on 80% of the days). In other words, washing our car was not associated with whether or not it rained.

The confirmation bias is caused, in part, by our unconscious tendency to ignore, avoid, or distort information that would show a preconception to be wrong. In the present example, people tend to pay attention only to the first cell of the table and to ignore the rest. This is because, in general, we are much more likely to notice when something happens than when something doesn’t happen. By forcing ourselves to pay attention to all relevant information in such situations, we are more likely to realize when our preconceptions are inaccurate.

I_Want_To_Believe_01

To summarize: Cognitive researchers have found that we have an automatic (unconscious) tendency to seek out and readily accept information that agrees with (confirms) our preconceptions, and to ignore, distort, or discount information that contradicts (disconfirms) them. This confirmation bias serves to maintain and strengthen our preconceptions: we are much more likely to perceive and remember experiences that confirm our prior beliefs, and to discount or reinterpret those that disconfirm them. Thus, over time, the confirmation bias results these beliefs becoming so well established in our minds that eventually we consider them to be common sense (i.e., obviously true). If we wish to minimize the effects of the confirmation bias, we must force ourselves to look for and examine closely both confirming and disconfirming evidence.

Read Full Post »

Boing Boing‘s Cory Doctorow reported on a spurious correlation of nearly 1.0 between autism and organic-food sales discovered by Jasonp55 on Skeptic Reddit.

This inspired me to look for other extremely high (albeit spurious) correlations with autism. I discovered a correlation of 0.994 between college costs (tuition + fees) and autism rates between the years 1999 and 2007, inclusive.

In the article that I’m certain to get published in Science, my main conclusion will be this: if we want to slash autism rates, we’ll need to drastically reduce college costs by returning educational funding to the levels of previous decades.

Here’s a graph of the cumulative percentages of the two variables that shows clearly their close association.

College-Autism

You may contact me at drjeffryricker@gmail.com

Data Sources
1. Office of Special Education Programs, Data Analysis System (DANS), OMB# 1820-0043: “Children with Disabilities Receiving Special Education Under Part B of the Individuals with Disabilities Education Act”
Table 1-11. Number of children and students served under IDEA, Part B, in the U.S. and outlying areas by age group, year, and disability category: Fall 1999 through fall 2008 (Age Group 6-21)
http://archive-org.com/page/2071756/2013-05-12/https://www.ideadata.org/TABLES32ND/AR_1-11.htm
2. National Center for Education Statistics: Digest of Education Statistics 2010 Tables and Figures
Table 345. Average undergraduate tuition and fees and room and board rates charged for full-time students in degree-granting institutions, by type and control of institution: 1964-65 through 2009-10
http://nces.ed.gov/programs/digest/d10/tables/dt10_345.asp

Read Full Post »

%d bloggers like this: