| Author |
 Topic  |
|
|
viggozhang
Belgium
10 Posts |
 Posted - 12/21/2010 : 07:28:03 AM
|
Origin Ver. and Service Release (Select Help-->About Origin):
Operating System:
Hey guys, an urgent question here about log Gaussian fitting for your kind help. The case is like this,
Quite some data points. At first I take "ln" of the variable "x", and then plot "f(x)" vs "t=ln(x)" on a linear scale, i.e. horizontal axis of t, and vertical axis of f(x). Afterwards, I fitted the curve by Gaussian f(x)=y0+2A*exp{-2(x-u)^2/w^2}/w(2Pi)^2 in Origin. (I noticed the definition difference of "w" in Origin, and which is not what I wanna ask.)
So here come my questions,
1. In this case, if the fitting is quite Ok, does it mean that I can call my data distribution "Log Gaussian"?
2. I think I made a mistake in the above-mentioned fitting procedure. So I gave it a second try: I multiplied my "f(x)" by "x" first, and then did exactly the same thing for fitting "x*f(x)" vs "t=ln(x)". The fitting is quite Ok. I think I can only call this a log Gaussian fitting and then use the corrected expressions for "u" and "w" in stead of using "u" and "w" values directly.
I know this is a relavtiely simple question, but I am really getting killed by this problem =.= maybe not smart enough... so, SOS~~
|
|
|
larry_lan
China
Posts |
Posted - 12/21/2010 : 10:44:10 AM
|
Are you talking LogNormal distribution?
http://en.wikipedia.org/wiki/Log-normal_distribution
It's not simply Log the independent data. Besides, Origin has a build-in LogNormal fitting function, just in Origin Basic Function cateogry. Maybe you can have a try?
If you transform your x, it just means data follows Gaussian distribution in that space.
Thanks
Larry |
 |
|
|
viggozhang
Belgium
10 Posts |
Posted - 12/21/2010 : 11:52:09 AM
|
larry, many thanks. yup, I meant LogNormal. Shooooot, bad eyes make life that bad... I got it.
Merry Christmas.
|
|
|
| |
Topic |
|
|
|
SOS: Log Gaussian fitting