with kernel K and bandwidth h [10], where kernel K may
be a suitably general probability density function, typically
unimodal and symmetric about zero, centered right over each
data point. The influence of each data point is thus spread over
its neighborhood, and the contribution from each point is
summed in the overall estimate. Bandwidth h is a scale factor
controlling how probability mass is spread around each point,
besides smoothness - or roughness - of density estimate, over
which it exhibits a strong influence.