Lies, Damn Lies and Statistics

Yesterday was my 1000 post, according to Blogger which keeps track of such things.  The little map I have on my blog site to the right and about half way down says I have had 46,729 hits since I started.  Since it seems to count whenever I go there too, I would round it back to about 45,000 hits and call it good.  Monthly hits look like this:

 

I have 40 admitted followers. Readership is steadily increasing and the number of countries astounds me.  I didn't save the January stats yesterday but think there were at least 50 countries, most with one to three hits but several with 20 to 50. 

Which brings up the subject of statistics and its uses.  I am NOT a statistician, I am an analyst (self-taught).  We had a statistics department when I was with Saskatchewan Agriculture.  Statisticians make me mental.  They are so precise, worse than accountants.  Accuracy and consistency are paramount.  Even if the numbers are not 100% (e.g. GDP or July 1 Cattle on Farms), the methodology must be explicable and consistent. And it is a good thing it is so.

Their job was to provide as accurate data as possible to the likes of me.  My job was to squeeze management information out of it.  Stuff you could make decisions from.  About and approximately were good enough.  So I would round it, massage it, reorganize it sixteen ways to breakfast to see what I could learn or to demonstrate a point or substantiate a policy position.

Statistics don't lie IF they are honestly gathered and tabulated.  To me, anything less than that is unethical and unprofessional on the part of the statistician.  It is what is done with them that makes the difference.  The analysis, the spin, the presentation, what is left in or left out, the conclusions drawn.

Just a simple example.  I could have divided the monthly hits by the number of days and got daily hits.  I could have looked at seasonal variations and blog posts per month variations.  I could have (if I had saved each months detailed stats) reworked the data excluding hits from countries with less than 10 hits. Or just looked at Canada and USA or even just Canada.  What might I have learned?  I don't know.  Maybe that increased readership in America accounts for the majority of the total increase.  What would that mean?  Darned if I know.

Look at the chart again.  The X axis is perfectly horizontal and the Y axis at perfect right angles to it, so the slope on the line is truly as it appears.  I could have (if I knew how) sloped the X axis upwards somewhat, keeping the Y axis perfectly vertical, changing the perspective from which you view the chart.  The information on the chart would still be perfectly accurate BUT the slope of the data line would appear steeper than it actually is.  The impression left with the viewer is that the number of hits is rising much faster than it actually is, unless the viewer takes time to read the chart carefully.

But any and all of this is only possible if the original statistics are accurate to start with.