The naive kernel density estimator definition

262 Views Asked by At

On B W Silverman Explanation of KDE:

It is stated that from the definition of a probability density, if the random variable X has density f, then

$f(x) = \lim_{h\to 0} \dfrac{1}{2h}\Pr(x-h < X \le x+h) $

And that a natural naive density estimator is

$f(x) = \dfrac{1}{2hn}[\text{no. of } X_1,..,X_n \text{ falling in }(x-h, x+h)] $

Where is the $\frac{1}{2}$ coming from?

If I have 2 values from a sample of 10 on the $(x-h,x+h)$ range $\Pr(x-h < X \le x+h) = \frac{2}{10}$ how does multiplying by $\frac{1}{2h}$ gives me the density?

1

There are 1 best solutions below

1
On BEST ANSWER

The probability $2/10$ is the area under the density from $x-h$ to $x+h$, which has length $2h$ and has height $f(x)$ when approximating the area by a rectangle, thus

$$ \frac{2}{10} = f(x) \, 2h $$