Sunday, October 19, 2014

Autism Rates and Movie Production Costs: Connection?


There is certainly no shortage of attempts to explain the rising prevalence of autism. I would like to present another. As we know, the rate of autism in 1997 was reported as being 1.6 per 1,000. That has sharply risen to 6.6 per 1,000 in 2007, and is currently reported as 14.7 per 1,000 in 2014. By comparison, the cost to make and market a movie in 1997 was around $60 million, increasing to $100 million in 2007, and $180+ million currently. The cost has more than tripled in just 17 years, coinciding with the sharp increase in the prevalence of autism.

It is likely that there are other factors involved, as the correlation does not appear to be linear. In addition, 1942's Casablanca cost just over $1 million to produce. During that same year, the rate of autism was effectively zero. This suggests that there may be a safe level of movie production cost, without risking elevating the autism rates.

It is not known at this time exactly how these two might be related. Some possibilities include more realistic special effects directing children to focus too much on small details, or more aggressive marketing strategies may lead to obsessive interests. Whatever the cause may be, I think further research may be merited.

Odds are very good right now that either you're laughing hysterically or you think I've gone completely nuts. In truth, I created a deliberately ridiculous example to make a point about correlation and causation. I don't believe that any sane person is likely to believe that movie production costs have anything to do with autism. However, there are many studies and theories that have little to no more validity than this, including several that gain widespread media attention. Unfortunately, most people are not adequately equipped to recognize a bogus or unusable study when they see one. I'd like to help by providing a quick guide to understanding the experimental process.

To start, I feel I should point out one of the basic rules of logic: correlation does not imply causation. Just because two things are happening at the same time does not mean they have anything to do with each other. They might be related, but further evidence will always be needed to establish a connection. My hope is that point is made above in the opening of this piece.

Second, it is important to always clearly define your terms. This is often overlooked in autism research, and perhaps other psychological studies. As definitions change over time and methods improve, researchers often neglect to adjust previous findings to account for these changes, leading to faulty results. If you look carefully at the above example, you may find a result of this fallacy. (Hint: Look up when Leo Kanner first published his autism research.)

The primary purpose of any experiment or study is to isolate one particular variable as much as possible. If there are too many variables, it can be difficult to say the reason for the results. There are multiple ways to isolate one variable. As many of them as possible should be used.

The first is to use a large sample group. Coincidences and unusual phenomena happen. With a small enough sample group, it can be difficult to tell if you're looking at a coincidence or an actual result of the experiment. The sample group should be large enough that coincidences should be expected. By doing this, it becomes easier to tell if a particular occurrence is happening at a statistically significant rate.

Next, we have to account for individual variation between different people. The best way to control for this is to have a diverse and representative sample. Many autism researchers look to particular programs or classes for their research subjects. The problem with this is that many of these programs only accept people with a defined age or level of intelligence, functioning, or income. Most of them also largely constitute people within a certain area. This means that autism is no longer the only variable involved in the study. The findings become unusable outside the one type of group in the study.

The third way we have to isolate a specific variable is to use a control group. A control group is a second sample group, similar to the experimental group, but without the variable being tested. The purpose is to observe alongside the experiment what a normal result would look like.

One type of control that's commonly used in treatment studies is a placebo. Often, a person will react to a treatment just by virtue of expecting a result. A placebo is designed to separate this phenomenon from actual results. The way it works is the subject is given an inert equivalent to the treatment and is allowed to believe that it is the actual treatment. Only differences between the actual treatment and the placebo group should be noted.

Another technique is a blind or double blind study. A blind study is when the participants do not know whether they are the experimental group or the control group. Most studies are conducted this way. A double blind study is when the researcher present with the participants also does not know which group is which. This is done to prevent the researcher from giving subconscious clues to the participants, or from subconsciously biasing the results.

To illustrate the isolation of variables, I'd like to use what may be considered to be an extreme example, Andrew Wakefield's research into vaccines. I realize this is a controversial example outside the scientific and medical communities. Allow me to explain why his research was never fully accepted.

First, Wakefield used a sample group of only twelve children. It is almost impossible to distinguish actual results from a coincidence with such a small group. Second, all of the children were drawn from his existing gastroenterology work, meaning that they all likely had gastrointestinal problems. That makes his results useless in terms of anyone with no gastrointestinal problems alongside autism. And third, he used no control group to compare his results. It is entirely possible that by using one or more control groups, he may have noticed similar results in unvaccinated autistic children, or different results in vaccinated neuronormal children. Either one of these would have rendered his observations irrelevant.

On a final note, you will occasionally find a study where the researcher appears to have decided on the conclusion before conducting the research. Usually, this is unintentional. Researchers are, after all, human like the rest of us. Sometimes, though, the data may be cherry picked to suit the desired conclusion. If you look at my example of the costs of movie productions, an astute reader may note that those costs have more or less risen steadily over the course of time, while the rates of autism diagnoses appears to have sharply risen in the 1990s, shortly after Hans Asperger's research was translated o English.

The check for this is usually the assumption that science is repeatable. If other researchers are unable to reproduce the same results, there is probably something wrong with the study.

If I may return to the Wakefield study into vaccines, there are two solid reasons to suspect his results may have been influenced by his desired conclusion. First, as often as his theory was put to the test, no other researcher has ever been able to reproduce his results. Second, it was later proven that at least five of his twelve subjects were showing signs of autism prior to vaccination. This second point, in particular, strongly suggests that this study may have been fraudulent.

The major take home message from this should be to not believe everything you read on the internet. There are certainly plenty of valid studies out there. Just be sure to check the methods and double check the data before you believe it.

No comments:

Post a Comment