regarding which variables/parameters/effects are important and which can be safely ignored
Many of the potential variables don't affect the results: the value of q, the continuum level chosen, the wavelength range over which the fitting is done, the inclusion of the background spectrum. A few of them do, though:
- The inclusion of the HEG data can make a significant difference. The peak line fluxes are lower by a factor of two for the 15.014 line, but still, using the HEG data does provide tighter constraints. It seems that we should use the HEG data when its flux level is within a few of the MEG data for a given line.
- The value of beta makes a difference in the derived model parameters. The cost of using non-integer beta isn't that high, so if there's good reason to think its value differs from unity, then it should be used. However, if a value of beta=1 is just as likely as 0.9 or 0.8, then model calculations can be done more rapidly with that value.
- The derived optical depth value (and, to a lesser extent, the value of uo) is quite sensitive to the assumed terminal velocity. As this value isn't known precisely, and the dependence of taustar on it is systematic (taustar is anti-correlated with vinf), we should probably test different values and report the results of fitting with different assumed terminal velocities.
- Finally, the choice of opacity bridging law in models with porosity can have a significant effect (on the derived values of the porosity length, though the profile morphologies and fit qualities are quite similar). As with the terminal velocity dependence, it's not clear which bridging law to use, so we should perform fits with both.
Now, some of the parameters that didn't have a big effect on this particular line could be more important for other lines. We should, for example, test the effects of letting q be a free parameter for some other lines (where, maybe, the best-fit value of this parameter will differ significantly from zero). We also haven't tested it in conjunction with porosity, for example. Continuum and background could be more important for weaker lines, and these could be checked in at least one other case.
Back to main page.
last modified: 25 April 2008