Discussion: sigma scaling factor

[adeane-ga]

This is just a question I have up for discussion: What do we mean when we say 2 Sigma for the error of the velocity estimates in PyRate?

Summary

• PyRate uses scipy.stats.linregress() to calculate the velocity (or slope of the trend line). It returns the standard error of this slope estimate. https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.linregress.html
• This results in the per pixel values in the linear_error.tif output file in the velocity_dir.
• PyRate has a config parameter called velerror_nsig which determines whether we report 1 or 2 "sigma".
• The default is 2 where it literally takes this input argument and scales the standard error of the velocity.
• Should we be using 1.96 to reflect the formal definition of 95 % confidence?
• Or does this not matter?

[mcgarth]

Assigning @richardt94 for comment, who implemented this feature

[adu461386118]

Hi Al, these are some lectures slides in Probability and Statistics from UTS about the 95 % confidence, personally, I think people tend to use 2 instead of 1.96 is because is it easier to do the calculation,