Computation of noise stats should use dtype=np.float64

[david-macmahon]

Computing the mean and std of large frames (e.g. ~16 million points) can suffer from floating point error when they are performed using float32 values. Both mean and std can be passed dtype=np.float64 to perform these computations using float64 instead and thereby retain better accuracy. One place where this could be added is here:

setigen/setigen/frame.py

Line 398 in 5ca049f

self.noise_mean, self.noise_std = np.mean(clipped_data), np.std(clipped_data)

but perhaps it would be good to add a dtype=np.float64 parameter to the enclosing function so that the user could specify a different type if desired.

[david-macmahon]

There may be other places where this change could be applicable as well.