Dear Uncle Colin,

According to John D Cook, you can estimate the natural logarithm of a number x by working out log2(x)log10(x). It seems to be pretty accurate - but why?

Not A Problem - Interesting Exponential Riddle

Hi, Napier, and thanks for your message!

That is indeed a conundrum: why should ln(x)log2(x)log10(x)?

The simplest approach I can see is to apply the change of base formula: loga(x)=logb(x)logb(a). In particular, log2(x)=ln(x)ln(2) and log10(x)=ln(x)ln(10).

This gives us a right-hand side of ln(x)(1ln(2)1ln(10)). If that bracketed factor is about 1, then the approximation is good.

And, it turns out, it is good: 1ln(2)1ln(10)=ln(10)ln(2)ln(10)ln(2)=ln(5)ln(2)ln(10), which we can rewrite as log10(5)ln(2) - because those are numbers you cover in week one of Ninja school 1

  • ln(2)0.69315
  • log10(5)=1log10(2)10.30103=0.69897

Those numbers are not identical, but they’re pretty close together: the difference is about six parts in 700, slightly under 1%.

Hope that helps!

- Uncle Colin

* Thanks to @pozorvlak for pointing me at John’s article.

Footnotes:

1. Or so I’m told.