Reliability of fit

[Timothy Kinney]

Fityk folks,

I like Fityk and am planning to use it for some FTIR data I am
publishing this summer. However, I need to convince my adviser that we
can trust the fits it generates. One thing I find problematic is that
if I perform the following set of functions, I get different peak
areas for the same data:

1) Load data
2) Restrict range to peaks of interest (in this case one peak)
3) Manually select a baseline
4) Subtract baseline
5) Add peaks (Gaussian)
6) Fit
7) Revert (to see peak area plotted with spectrum)
8) Export Peak Parameters

When I do this multiple types for one data set I get peak areas
ranging from 4.63 to 6.21, which is a pretty significant difference
consider it is the same data and fitting algorithm. I think the
biggest problem is probably with my manual selection of the baseline.

Is there a way to specify the baseline more reliably each time in
order to improve the reproducibility of fits? Or am I doing something
else wrong?

Thanks for your time.

-Tim