Lately, I’ve been noticing that people are using statistics more often to lie instead of using common sense.
For instance, I work with a friend who was very upset because it wasn’t recognized that she increased survey participation by 50% versus the prior month’s participation numbers. Since I’m a little more mathematically astute, she came to me to check if her calculations were correct (and to complain). So I quickly ran the numbers in my head and she was correct, she did increase the survey participation by 50%.
Now most of us would agree that a 50% increase is fairly impressive, at least from a mathematical perspective!
Statistics Don’t Tell The Entire Story
The catch with my friend is that the number of participants involved with the survey the prior month was only two. The month that she became involved with the survey results, the number increased to three participants. So while technically she can claim that her involvement increased the participation rate by 50% versus the previous month… it’s still just one additional person!
To really make matters worse was the fact that her manager was upset because 2 month prior the survey participation number was four! So in her manager’s viewpoint, the numbers actually decreased by 25%. While I wholeheartedly agree with my friend, when you speak in terms one or two people, the statistics really doesn’t matter much. It could be just luck that one more person filled out a survey in her month than versus the previous month, or two months ago.
Obviously, she was right in her calculations, but the significance of one person doesn’t really matter. Especially when the goal is for having hundreds of responses per month instead of just single digits numbers.
I personally find that a floating average number, or some other kind of baseline is the best for comparing performance.
For example, with my friend, if the average participation rate for the surveys is one person per month, then realistically 3 actually is a phenomenal number and she should be acknowledge for her efforts.
I’m writing this because all to often I see people on all sides of arguments using mathematically correct statistics in a way that promote their argument without taking into account the historic average of numbers the statistics represent. This is a common tactic with politicians in general.
Statistics that Use Bad Sampling Sets
Okay, I’m going to get nerdy on everybody here, so just bare with me… If the sampling set taken isn’t representative of the general population or the target population being represented, then the statistics that are used on the non-representative sample will be inaccurate. For example, if you take a count of the eye color of 20 Swedish people as your sample set, you’ll derive a number that states that % of the eye color of everybody is blue. Obviously, we know this isn’t true, but such sampling occasionally happens, especially in politics. The above picture declares Dewey is the new president, but we know he wasn’t elected. This was due to an error in the sample set or size of the sample set.
What to Use When Statistics Lie or Are Abused
Try to find unbiased results. This isn’t easy though, since usually the statistics that a person or group presents was collected by them for the presentation. This alone should set off red flags! How accurate can the statistics be if they are using the statistics for their presentation? Wouldn’t be be kind of silly of them to present statistics that would undermine there cause?
Personally, I try to come to a conclusion based on common sense with such matters. This isn’t easy and is highly subjective, but without conducting a fair and representative sample, it’s just not possible to get what the real statistics are around a topic.
Statistics that Don’t Account For All Variables and Time
There is a concept called Spurious Relationships, which basically exists when two variables seem related but actually aren’t. For instance, one of my favorite examples is that ice cream consumption increases in summer and so does the temperature in that given area. So does eating ice cream raise the temperature of the climate in summer? No, but yet the two variables seem to be highly correlated…
So basically, what I’m trying to say is just because a correlation may seem to be a statistical number that make sense, often time it doesn’t. Using common sense, you can debunk such poor statistical usage and you should question the presenters and their motives. Don’t be fooled by numbers, just because someone uses statistics doesn’t mean that they are right.
Don’t believe the hype, think things through!
MR