# Graphing 11/22/21

Deception with Richard Bleil

Recently a graph crossed my path that was intended to help raise funds.  The advertisement argued that the stock was skyrocketing, and indeed, it looked as if it was doing just that.  Showing a series of ups and downs, a sudden spike at the very end looked enormous.  But it wasn’t.

I suppose as an analytical chemist, I’ve been trained to look at the instrumental noise level.  Any instrument has a certain amount of random fluctuations in the detector, and if you zoom in sufficiently it looks pretty large, but it’s really just the zooming.  This is something that advertisers will do to deceive prospective clients.  Remember to look at the scale provided (and if there is no scale take it as an enormous red flag).  If the scale is very small, any fluctuations will look enormous.  If the scale is in the cents, you’re not looking at dollar values.  As an analytical chemist, if we’re looking for peaks (indicating the presence of a compound), we discount any peaks unless it is at least three times higher than the noise level.  In other words, if that same graph is fluctuating from two the seven cents, nothing is significant unless it is around fifteen or twenty cents.

Auto-scaling graphs are a real eye-opener for students.  As they are watching a real-time analysis of an instrument, at first it looks like there are thousands of peaks as the scale is set such that the random fluctuations fill the entire screen.  When a real peak begins to come out, the system “auto-scales”, and suddenly all of those tiny little peaks are squashed down to nearly nothing as the first actual peak shows up.  The peaks that looked so significant at first suddenly look like nothing at all, just the baseline.

That’s not what this advertiser was doing.  That last peak still looked very large, which was so unusual compared to the rest of the graph that any wise investor wouldn’t take it seriously as it was likely a random bit of data, called an “outlier” by statisticians.  These are data points that don’t really match the remainder of the data.  They can be caused by any number of reasons.  In investing, for example, maybe somebody is trying to buy up all of the stock they can and are willing to overpay for it.  It’s not necessarily a wise approach, but once they’ve bought all they want, that stock is likely to drop back down again.

I purchased stock in a pharmaceutical company just before their Coronavirus vaccine hit the market.  I expected the stock to skyrocket, and indeed it did, to about three times more than what I had paid for it.  As news of the vaccine hit, the stock fluctuates accordingly.  As its efficacy is demonstrated, it shoots up.  As news of side effects hit, it drops.  As news of a secondary pandemic hits, it skyrockets.  It’s like a roller coaster, but despite the fluctuations, it’s still improving.  Since purchasing it, it has lost most of its value (as of the writing of this blog).  Today it’s worth only twice what I paid.  It lost half of its value compared to the peak, but I’m not planning on selling, because despite the fluctuations, overall, it’s going up.

But that’s not what this advertisement did either.  No, this advertisement actually did something that is very common in statistics, namely, reset the baseline.  See, the graph’s base was not set to zero.  The baseline was quite high (I don’t have an exact number).  The graph was set to the low point in the fluctuations rather than zero.  This has the effect of making those relatively small fluctuations seem enormous.

For example, let’s suppose that over the course of three months, a stock value fluctuates from a low of, say, \$198 a share to a high today of \$212 per share, then on a graph, that sudden peak will look huge when, in reality, it’s only \$4 and probably only around 2% of the overall value.  That’s hardly a rally, but it would fill the screen if the graph was between \$195 and \$215.  Always look at the base value.  And while we’re at it, don’t only look at the y base, but the x as well.  How much data are they giving us?  A couple of years?  Or a few hours?  This makes a difference as well.

It’s been said that with statistics you can prove anything.  That’s pretty much true, and many of us feel comfortable interpreting graphs because, if you are like me, graphing was pretty much my best subject in math.  This gives a false sense of security when looking at a graph that suddenly shoots up, convincing me that I need to buy quickly before it becomes cost prohibitive.  Honestly, it may already be so if I can’t afford \$200/share, and if I can, maybe the rush isn’t as great as the graph seems to indicate.