Saturday, December 18, 2010

Lying with Statistics...and PowerPoint

It is well known and obvious to most of us in our profession that the phrase "lying with statistics" does not mean to be intentionally misleading. In fact, knowledge of the ease at which someone can mislead with the manipulation of numbers is common enough to prompt at least a cursory review of most analysis products to head off the inadvertent case of manipulation that could be construed as intentional deception. That is of course a cover your butt response to head off any low hanging criticism. But it is an important one. It is not the altruistic goal we seek of conveying only the truth. Nevertheless, the early discovery of that which may be misleading does have a certain first order effectiveness. It helps by eliminating early distractions as we should always strive for our products to be clear and as absent of controversy as possible.

This doesn’t correct the problem, however, if the result being reported is, itself, incorrect because the analysis behind it was faulty. Still yet we must be reminded that it is extremely important in our search for the truth that we understand how it is very easy to lie with statistics and that we must always be on guard to keep deceptive or misleading results completely out of our activities. Further, we must avoid the unintentional pitfalls of the of graphical products such as PowerPoint that obscure our message, change our message, or become the message...because they will.

For today this essay is not about the mechanics of lying with statistics, the well known book, “How to Lie with Statistics” by Darell Huff covers this topic quite well. On the issue of Power Point, the books and publications about graphical presentations by Edward Tufte, such as “The Graphical Display of Quantitative Information”, pretty much cover the water front on that topic as well. Since in both cases, graphical presentation of statistics and PowerPoint are primary tools in our tool kit and will remain at our immediate disposal for sometime to come, what more can be done to extract truth?

The best way to avoid the pitfalls is to keep your perspective that you are the analyst. Tools and presentation products do not have the brains – you do. Long before you commit a number to a spreadsheet you should know exactly what trend you are trying to show or display. If you find yourself manipulating the data in the spreadsheet because you are looking for meaning in the data, you probably are beginning to walk on thin ice and should be wary. If the trend doesn’t just jump right off the page it’s possible that it’s not really there. This should be your first signal to start questioning what came before the spreadsheet, not what’s in the spreadsheet.

With regard to briefing slides, here again, you should have a pretty good picture in your mind with regard to the story you want to tell. If you let the story emerge while constrained by the software application, you are probably letting something creep into your message that you didn’t intend. When Marshall McLuhan said, "The Medium is the Message", he was serious about the underlying facts and content not being the message. The message is contained in the medium. Over and over again we find this to be the case. A well presented diagram of the facts trumps the facts in most cases. What this means if if you present a beautiful depiction of the relationship between two sets of data, it is the picture of this relationship that will be remembered in the mind of the decision maker. You better hope it was a correct relationship and not a misleading relationship. Time and time again we see presentations that show a wide gap between two alternatives. If the scale was changed, and the observer is not aware of the change in scale, they come away with a vision of separation that may or may not actually exist...the medium became the message. And the message was not necessarily true.

The best way to combat this is to decide on the truthfulness of your message first. This means do the thinking first and then use the tools at your disposal, not visa versa. In this way, the truth, as you will learn to discover, will not be obscured, influenced, or manipulated –either intentionally or unintentionally, by the external limitations some of which you were not even aware. Then your final product will be clear and logical and stand-up to the first wave of PowerPoint critics who enter the room...and those who leave the room will not be deceived.

2 comments:

  1. Another interesting read on the topic of misrepresenting statistics (and the underlying data) is "Sex, Drugs, and Body Counts: The Politics of Numbers in Global Crime and Conflict" (http://www.amazon.com/gp/product/0801476186/ref=oss_product). The book is actually a series of journal articles on topics such as human trafficking, war casualties and money laundering, to name a few. The editors, Peter Andreas and Kelly M. Greenhill, take issue with the lineage of the data presented and analyzed -- it's usually false and extremely difficult to measure. The problem: we are still reporting these figures and statistics in the media; we are still creating policy using, in many cases, grossly inflated figures. I have a copy for those interested in borrowing it.

    ReplyDelete
  2. This does look interesting I will check it out.

    ReplyDelete