weighted_operations

jwst.ami.leastsqnrm.weighted_operations(img, model, dqm=None)

Perform least squares matrix operations to solve A x = b weighting by Poisson variance.

A is the model, b is the data (image), and x is the coefficient vector we are solving for.

Here we are weighting data by Poisson variance:

x = inv(At.W.A).(At.W.b)

where W is a diagonal matrix of weights w_i, weighting each data point i by the inverse of its variance:

w_i = 1 / sigma_i ^ 2

For photon noise, the data, i.e., the image values b_i have variance proportional to b_i with an, e.g., ADU to electrons conversion factor. If this factor is the same for all pixels, we do not need to include it here.

Parameters:

img2D float array

Input data

model2D float array

Analytic model

dqm2D bool array

Bad pixel mask, same shape as image.

Returns:

x1D float array

Coefficient vector

res2D float array

Residual; difference between model and fit

Notes

Use matrix_operations() for equal weighting of data.