I have two Gaussian curves, and I would like to sum them to have a curve with two bells, to fit some bimodal histogram.
Is doing N(m1 + m2; sig1 + sig2) the good way or should I do something else ?
For instance, I would like to obtain something like the green curve :
Thanks for the help !
EDIT: to fit my histogram curve, I have developed the EM algorithm, giving me back the best means / sigma for the laws, and also laws coefficients, but I don't really know how to use these coefficients to build my mixed curve.
$N(\mu_1+\mu_2,\sigma_1^2+\sigma_2^2)$ corresponds to the curve $\displaystyle x \mapsto \frac 1 {\sqrt{2\pi}\sqrt{\sigma_1^2+\sigma_2^2}} e^{-(x-(\mu_1+\mu_2))^2/(2(\sigma_1^2+\sigma_2^2))},$ and that has just one "bell", centered at $\mu_1+\mu_2.$
What you need is $$ w_1 \frac 1 {\sigma_1\sqrt{2\pi}} e^{-(x-\mu_1)^2/(2\sigma_1^2)} + w_2 \frac 1 {\sigma_2\sqrt{2\pi}} e^{-(x-\mu_2)^2/(2\sigma_2^2)} $$ where $w_1$ and $w_2$ are weights, i.e. positive numbers whose sum is $1$.