Unexpected values of the standard deviation of the residuals (Rev.St.Dev.)

55 views
Skip to first unread message

Walter

unread,
Oct 7, 2021, 10:56:00 AM10/7/21
to fityk-users
Hello everyone,
I’m finding some troubles using Fityk with the standard deviation of the residues (Res.St.Dev.). I need that information to compare some datasets. After a while I decided to generate the data with excel (knowing all the parameters) and import them on Fityk to compare them.
I generated a Gaussian signal s=10*exp(-(x^2)) with x 100 points from -2.5 to 2.5. Then I generated the noise with n=NORM.INV(RAND();0;DS) the value of DS were set to 0.1; 0.5; 1 and 5. The datapoints are the sum of the two: d=s+n. Calculating the standard deviation of the residual (the noise n) I obtained DS as expected.
Then I imported the datapoints in Fityk, using a guassian fit, and instead of Res.St.Dev. 0.1; 0.5; 1 and 5 I obtained 0.081; 0.371; 0.704; 3.643.
If I add a constant to the gaussian : s=10*exp(-(x^2))+100 the residual do not change but in Fityk (obviously inserting a constant also in the fit) the values of Res.St.Dev change a lot and are even worse : 0.0099; 0.0481; 0.1044; 0.5091. They are actually lower but farther than the real ones.
Anyone can help me to know why?
Thank you
Walter
Reply all
Reply to author
Forward
0 new messages