3

I was doing a probability question where it was required to find out the least integral value of n with $$f(n)=\left(\frac 9 {10}\right)^n<\frac{1}{2}$$

A little calculation(by calculator of course) yields $\min(n)=7$.

as:

  • $f(5)=0.59$
  • $f(6)=0.53$
  • $f(7)=0.43$

My question is there a good way to approximately tell which integer $n$ can be without computational software and log tables (even $\log 2,\log 3$ etc.)vCalculus seems to be an unlikely option

4 Answers4

4

hint

For small $ X $, We can write

$$(1-X)^n\approx 1-nX$$

or, for more accuracy $$(1-X)^n\approx 1-nX+\frac{n(n-1)}{2}X^2$$ and

$$\frac{9}{10}=1-\frac{1}{10}$$

2

Rewrite what you want as

$$2\approx\left(10\over9\right)^n=\left(\left(1+{1\over9}\right)^9\right)^{n/9}\approx e^{n/9}$$

If you allow yourself to know that $\ln2\approx0.7$ (which is the basis for the "Rule of $70$" described in Rivers McForge's answer), then you get $n\approx9\ln2\approx6.3$, which accords with what you found.

Barry Cipra
  • 79,832
  • thank you so much! The other answers were not strong enough ie i still had to work out more cases.Could you also tyhrow some light to this question of mine https://math.stackexchange.com/questions/3886271/proof-for-that-exactly-one-normal-can-be-drawn-to-a-parabola-from-an-exterior-po?I hope i am not troubling you – Albus Dumbledore Nov 05 '20 at 12:19
1

As $$ \left(\frac{9}{10}\right)^n \lt\frac{1}{2} \Rightarrow \left(\frac{10}{9}\right)^n \gt 2 $$ $$ \Rightarrow \left(1 +\frac{1}{9}\right)^n\gt 2 $$ Bernoulli's Inequality shows $n=9$ can be the solution.
Now we start descending from $n=9$ to $n=8$ and $n=7$ and eventually see $n=6$ does not suffice, hence the solution is $n=7$.

  • Hmm thats nice too, but the problem is calculating (1+1/9)^9 is going to be tedious.Any way it is still more efficient than random brute force(+1) – Albus Dumbledore Oct 31 '20 at 17:14
  • @AlbusDumbledore The thing is that we don't need to calculate $n=9$! That is what Bernoulli's Inequality shows, so we start from $n=8$. – Book Of Flames Oct 31 '20 at 17:17
  • yes i missed that but even n=8 is a little ugly. Again i think there is a limit to all these approximations .. – Albus Dumbledore Oct 31 '20 at 17:19
1

The informal heuristic you're looking for is called the "Rule of 70". In its most typical form, it says that if a quantity $A(t)$ increases by a percentage rate $r$ every unit of time, or $$A(t+1) = (1+r) A(t),$$ then the time it takes to double is approximately $70/r$. It can also be used for constant percentage decreases: if $$A(t+1) = (1 - r)A(t),$$ where $r$ is the percent decrease per unit of time, then the "half-life" of $A(t)$ (i.e. the time $\lambda$ it takes for $A(\lambda) = A(0)/2$) is again approximately $70/r$.

In this case, you have a constant percent decrease of $r = 0.10$, or 10%, so the Rule of 70 predicts the halving time should be $$70/10 = 7,$$ which is exactly what we found by checking each case.

  • It’s probably best to let $r$ be the percentage rate, as in the linked article. In that case, the formulas should be $A(t + 1) = (1 \pm \frac{r}{100}) A(t)$. – Brian Drake Nov 06 '20 at 14:15