Sunday, May 27, 2012

Exaggeration is ultimately counterproductive

When I teach stats, we discuss a bit how numbers can be used to deliver alarmist messages. Using a generic Compound X that shows no toxic effects up to 100 units with single digit baseline levels, we discuss how varying increases in X might be reported.  So let's say baseline levels of X are 2 units.  Something is done, and levels of X rise to 3 units and this change reaches the level of statistical significance.  The alarmist reporting that gives the greatest impact would probably be to state something like:  "levels of X increased by 50%".  This is true, and it sounds daunting taken out of context.  Alternately you might see something like "levels of X were 1.5 times baseline", though I think the 50% number has more impact.  For anything less than 10X increase, the alarmist reporting almost always uses the percents, because over 100% has emotional intuitive impact.  So for example if X values rose to 4 units, "X values doubled" has impact, but "increased by 100%" or even "were 200% of baseline" sound worse.  And let's' say the levels of X increased to 10 units.  Yeah "five times" sounds less dramatic than "500% of baseline" or "increased 400%", and definitely more dramatic than just presenting it as an "8 unit increase".   The reporter of such information trying to exaggerate the impact to drive an agenda does not tell the consumer of the information about X's safety threshold.  And chances this is not common knowledge this information consumer possesses.   But I contend that reporting data in this way is misleading.  If it's just some detergent manufacturer boasting stain removal or something like that, who really cares ... but if it's a scientist looking to influence public policy and regulation?  
Read more »

No comments:

Post a Comment