(The views expressed here are those of the author, the founder

of FridsonVision High Yield Strategy.)

By Marty Fridson

May 21 (Reuters) - One person you wouldn't expect to

hear tout a statistical fallacy is Warren Buffett, but the

legendary investor appeared to do just that at the recent

Berkshire Hathaway ( BRK/A ) annual meeting, a reminder of just

how easy it is to fall into statistical traps.

While speaking at the annual meeting - otherwise known as

"Woodstock for Capitalists" - in early May, Buffett made the

following rather odd comment while holding up a can of

sugar-laden soda:

"For 94 years I've been able to drink whatever I want to

drink. They predict all kinds of terrible things for me, but it

hasn't happened yet ... Charlie (Munger) and I never really

exercised that much or did anything - we were carefully

preserving ourselves for these years."

Buffett has a deep knowledge of the insurance business. So

we can be confident that he understands actuarial methods and

probability theory better than his remarks suggest.

But that might not be the case for everyone in the

conference audience.

While plenty of investors have prospered over long periods

without the benefit of formal training in statistics, an

inability to think probabilistically can lead to serious errors.

Understanding a few of the field's basic principles can help

everyone avoid the most common financial pitfalls.

STATISTICAL FALLACY

Assuming that Buffett's longevity somehow negates the

long-proven health benefits of diet and exercise is a clear

statistical fallacy. It's like pointing to a 90-year-old who

smokes three packs of cigarettes a day to disprove the

correlation between tobacco use and longevity. In reality, both

cases merely show that some people have exceptionally good

genes.

Correlations between health and bad habits predict the

percentage of individuals within a population who are likely to

have adverse consequences from engaging in these behaviors.

Those correlations predict nothing about the outcome for a

particular individual within the population.

It's much the same with investing. Just because one

company's stock defies a trend - like Amazon ( AMZN ) after the

bursting of the 1990s tech bubble - that does not mean other

seemingly overvalued firms will eventually become corporate

superstars.

Another common statistical mistake among investors is

blindly accepting a bearish prediction by a reputed genius who

successfully forecast a previous market decline.

If stocks suffer an especially large drop - proving the

supposed oracle "right" - the forecaster is apt to run video

clips of the dire warnings he flooded the airwaves with before

the down year. No mention will be made, of course, of the wrong

calls the "genius" made in prior years.

The fact is that the S&P 500 has fallen in 32% of the past

97 years. That means a forecaster can warn of an impending

selloff at the start of every year and pretty much expect to be

right one out of three times.

Of course, the one-out-of-three odds are based on averages

over a significant period of time, but the point remains that

inevitability, rather than insight, is a key contributor to the

fame of many forecasters.

STATISTICAL SIGNIFICANCE

Another way that unfamiliarity with statistical concepts can

lead to unwise investment choices is the failure to properly

understand statistical significance.

Let's say the stock of a fictitious company - call it XYZ

Corporation - generated the following annual total returns in

the 10 years after its initial public offering.

Comparing these results with the S&P 500's total return over

the same period, a broker might make a pitch like this:

"XYZ has outperformed the market for an entire decade,

posting an average return of 16.09% versus 14.15% for the index.

An advantage of nearly two full percentage points over the next

20 or 30 years would give your portfolio a gigantic boost over

the long term, thanks to the magic of compounding. These returns

suggest that XYZ's management team knows how to create superior

results for shareholders."

Kudos if your reflexive response is "Past performance is not

indicative of future results." But the problem with the broker's

sales pitch goes deeper. It asserts that XYZ's C-suite

executives have demonstrated managerial skill by engineering an

index-beating stock return. Missing from the discussion is the

critically important concept of statistical significance.

In non-technical terms, no genuine evidence exists that

XYZ's stock's performance edge was anything more than chance.

Confirming statistical significance would require an

understanding of several other quantitative tools, including

standard deviation, t-statistics, confidence intervals, and

p-values, as well as the use of a difference-of-means

calculator.

You'll be forgiven if you decide not to wander that far into

the weeds of empirically based financial analysis, but this

four-minute read on statistical significance would be well worth

your while.

In attending to your physical health, it's hazardous to fall

prey to statistical fallacies such as, "People who don't

exercise can expect to live as long - or longer - on average

than people who do. Just look at Warren Buffett."

Your financial health could be similarly jeopardized if you

fail to recognize that numbers presented in misleading ways can

lead incautious investors to make terrible decisions.

(The views expressed here are those of Marty Fridson, the

founder of FridsonVision High Yield Strategy. He is a past

governor of the CFA Institute, consultant to the Federal Reserve

Board of Governors, and Special Assistant to the Director for

Deferred Compensation, Office of Management and the Budget, The

City of New York.).

(Writing by Marty Fridson; Editing by Anna Szymanski and

Stephen Coates)