Standard deviation of intensities with a spectrum

[brechmos-stsci]

The standard deviation of the pixel intensities in the region, with optional sigma clipping (to use astropy sigma clipping).

The region could either be passed in or a sub-set of the spectrum extracted and then passed into this method.

[eteq]

As with #243 , I'm not convinced this should be implemented as a stand-alone function. Instead: np.std(spectrum.get_region(reg).flux)?

[eteq]

Reading this more carefully, what is the purpose of this, @brechmos-stsci ? I thought you meant the width of a line, but now I see this is about the flux. I wonder if this has been superceded by the uncertainty estimate thing in #248 ? If so I think this can be closed.

[brechmos-stsci]

Hmmm.. no... not the width. For all of these "stats" things, I took them to mean a method that operates on a spectrum or region. For this one, I took it as what you had in your comment, basically just a wrapper for np.std working on the flux. (A mean and median could be useful too).

The reason to actually implement these (rather than let the user do it) would be to have it in the package for consistency of interface.

[eteq]

Ah, so in that case I'm not really in favor of this. I think we don't want to add functions for consistency if there aren't science use cases driving them (so as to lower the maintenance burden). But if we encounter a significant science case that requires this (bearing in mind also that it's a one-liner to do by hand... but that could still be a documentation thing).

[eteq]

For completeness, in v0.4 that would be np.std(extract_region(spectrum, reg).flux)

[nmearl]

Closing in favor of #672.