error in help files? lognormalnoise
daggaz
Fri, 08/31/2012 - 04:18 am
1/(x*s*sqrt(2*pi)) * exp[ (ln(x)-m)^2/(2*s^2)]
shouldn't there be a negative sign just inside the exponent : exp [ -(stuff)^2] ?
As well, it seems pretty straightforward that m = mu and s = sigma. Can we get that listed explicitly? There are a few help files Ive noticed I believe where the variables are not necessarily defined like this.
Finally, and this is more of a wishlist kind of thing, wouldn't it be convenient for the user, if the help files/GUI dialogues and the noisefunctions/fitfunctions matched concerning choice of formula? Often when I am using a fit, I have to jump thru some math hoops to redefine the values in W_coef. One cant use gnoise, for example, with "width", one needs sigma. Having these relationships explicitely defined, or better yet, a button to quickly switch between the fitting parameters, would be very nice.
As for your request, there are two issues involved: first there is the question of backward compatibility (we can't change definitions of functions that have been in use for a long time) and second, just because a particular definition fits one's application (e.g., fitting) does not mean that it is ideal everywhere. A case in point is the Gauss function which returns a normalized distribution. As it turns out this was not ideal for curve fitting so IGOR sports Gauss1D and Gauss2D as well. Since Gaussians are used sufficiently often we added these special cases but the same argument does not hold for most functions.
A.G.
WaveMetrics, Inc.
August 31, 2012 at 10:23 am - Permalink