A new paper gives strong indication that on average, individual investors are really poor investors. I mean, really bad. Enough that by the end of the day, they’d be better off sitting in cash than trading on their own. It treats them to the sort of analysis which we usually subject mutual fund managers to, and finds that they’re no better, i.e. worst than a passive strategy. Continue reading Want to increase returns by 8.5%? Close your online trading account.
Fortunately, I was also exploring Understanding Uncertainty today as well, and came across this wonderful interactive illustration of exactly these laws. Great for seeing mortality rates in action (not often one says that…).
This ties in with my earlier post on how graphics, especially interactive ones, will help make statistical and probablistic analysis more mainstream and prevalent as more people are exposed to it. However, it does still require a sceptical and trained mind to understand the implications of sampling procedures, tests against the null etc. I do wonder if more easily available statistical methods (Google Analytics, SurveyMonkey’s graphing tools) will create a low hurdle for those with no statistical training to present “analysis” as if they do.
Historically, the mere fact you were able to put together a chart would identify you as a nerd. Now that statistics is the next sexy profession, everyone will try to jump on it. But just because you can make a chart doesn’t mean your analysis is meaningful or correct. I’ve seen more and more presentations by individuals who had zero training in statistics, but still presented “statistical evidence” by bar charts or pie charts (the hallmark of the statistically naive). My favorite was a presentation where a comparison of responses between two groups was done on count data – the fact that 200 (out of 2,000) respondents in Group A responded “yes”, while 40 (out of 200) in Group B responded “yes” was taken as evidence that Group A were more amenable to the change. Yikes.
Varian is correct in that:
The ability to take data—to be able to understand it, to process it, to extract value from it, to visualize it, to communicate it—that’s going to be a hugely important skill in the next decades… Because now we really do have essentially free and ubiquitous data. So the complimentary scarce factor is the ability to understand that data and extract value from it.
I think statisticians are part of it, but it’s just a part. You also want to be able to visualize the data, communicate the data, and utilize it effectively. But I do think those skills—of being able to access, understand, and communicate the insights you get from data analysis—are going to be extremely important. Managers need to be able to access and understand the data themselves.
But don’t underestimate the importance of that last sentence. In fact, that might be the most important sentence. The ability of a decision-maker to discern good, trustworthy statistical analysis from amateur hour will both increase the quality of decisions, and dissuade amateurs from attempting to use simplistic analysis to make their point. If the weakest link runs the show, the fact you’ve used propensity scores doesn’t matter.
Dan Goldstein notes the intuitiveness of using graphical representations of probability, especially in Bayesian settings.
High-quality , sometimes even interactive, graphics and charts, have increased greatly in recent years as news and other information have migrated to the web. The New York Times has a dedicated staff of graphical artists well versed in information design, and many digital graphical artists are becoming better versed in statistics and displaying ideas and results. Even the NHS in in on the act, reporting my surgery’s performance with graphs and relevant comparison groups. (Kudos!)
- Digital graphics are inherently tinkerable for the designer, reducing turn-around times on trying new displays and ensuring the message is correct.
- Probabilistic reasoning has become more important with the internet, more developed financial sectors, and more information. We thus have the need, and the circumstance, to use them more than our predecessors.
- Our eyes and brain are much quicker at judging sizes and distances than translating numbers to an equivalent internal representation, so we can read (good) graphs quicker. With more exposure, we read them even quicker.
- Many individuals unceasing and constructive criticism and leadership in creating better graphical standards.
- More people trained in statistics and related fields are moving into the public and private sector from academia, and trying to figure out how to communicate their ideas or conclusions to those without training.
- Will this affect existing foundations of behavioural economics? Much of the extant literature is derived from numerical representations of probability. Both the papers in Dan’s link are from within the past 10 years, whereas core papers in the decision science field (Kahneman and Tversky etc) are from the 1970’s. Gerd Gigerenzer’s work actively works at finding when such biases are solely based on presentation effects, in effect eliminating them with graphical representations.
- If graphical representations are consistenly found to enable superior probablistic reasoning, why aren’t they more widespread? I see far more cases where the producer (private business) loses from clear intuitive representation of charges, competition, and risk than benefits. A clear graph showing a specific business is the best only benefits one business.
- The example used it clear for bayesian categorical reasoning – but what about probability distributions and raw numbers? We ran an experiment a while ago with discrete probability distributions which did pretty well, but I’m sure there are futher applications.
- What’s next? Interactive versions such as Dan’s distribution builder is just one possibility.
For some truly fantastic graphics, which aren’t necessarily probablistic but do increase readability, see http://www.style.org/