Find upper limit of normal distribution integration

931 Views Asked by At

Considering the normal distribution with standard deviation equals to 0.9 and mean 2.1:

$$ P(X\leq a) = \frac{1}{0.9\sqrt{2\pi}}\int_{-\infty}^{a} e^{-\frac12\frac{(x-2.1)^2}{0.9^2}}\,dx $$

I must find the value for 'a' that satisfy $P(X\leq a) = 0.1$ using the numeric method Simpson 1/3 without repetition.

To simplify I named $c = \frac{1}{0.9\sqrt{2\pi}}$ and $f(x) = e^{-\frac12\frac{(x-2.1)^2}{0.9^2}}$

So $P(X\leq a) = c \int_{-\infty}^{a} f(x)\,dx$

Therefore $c\int_{-\infty}^{a} f(x)\,dx = 0.1$ and $c\int_{-\infty}^{2.1} f(x)\,dx = 0.5$.

To be able to use Simpson 1/3 I must have a integral with finite limits, so I write:

$$ c\int_{-\infty}^{2.1} f(x)\,dx = c\int_{-\infty}^{a} f(x)\,dx + c\int_{a}^{2.1} f(x)\,dx $$

$$ 0.5 = 0.1 + c\int_{a}^{2.1} f(x)\,dx $$

$$ \int_{a}^{2.1} f(x)\,dx = \frac{0.4}{c} $$

Using Simpson 1/3:

$$ \int_{a}^{2.1} f(x)\,dx \approx \frac{h}{3} (f(a) + 4f(a+h) + f(2.1))$$ with $h = \frac{2.1 - a}{2}$.

Solving for $a$:

$$ \frac{h}{3} (f(a) + 4f(a+h) + f(2.1)) = \frac{0.4}{c} $$

I get $a = -2.881562$.

Using math software, I know that $a$ value is wrong. But I don't know where I made a mistake.

So, how can I calculate the right approximation for $a$?

1

There are 1 best solutions below

0
On BEST ANSWER

[SOLVED]

Let $$ F(x) = \frac{1}{0.9\sqrt{2\pi}} e^{-\frac12 \frac{(x-2.1)^2}{0.9^2}} $$

And

$$ f(a) = \int_{-\infty}^{a} F(x)\,dx - 0.1 $$

We must solve $$f(a) = 0$$

$$ f(a) = \int_{-\infty}^{2.1} F(x)\,dx - \int_{a}^{2.1} F(x)\,dx - 0.1$$

$$ f(a) = -\int_{a}^{2.1} F(x)\,dx + 0.4$$

For that, a method like Newton's Method could be used:

$$ x_{k+1} = x_k - \frac{f(x_k)}{F(x_k)} $$

Since $ f'(x) = F(x) $.

Using Simpson 1/3 the integral in $f(a)$ can be numerically approximated.

In this conditions, using $x_0 = 1$ the fourth iteration of Newton's Method gives:

$$ x_4 = 0.948042356255413 $$

This result, clearly carries the error of approximating the value of $f(a)$, as Simpson 1/3 method has an error of about $|0.0012|$ for this function, using two intervals.