Bootstrapping error bars for spline derivatives
Resample the data with Gaussian noise of the quoted error, refit the smoothing spline, and take the spread of the derivative over the resamples. Keep the smoothing factor fixed across resamples.
Field Notes
Small notes on numerical methods, plotting and everyday tooling.
Choosing axis ticks that read well
Pick a step from 1, 2, 2.5 and 5 times a power of ten so that four to six ticks cover the range, then pad the limits by half a step.
Keeping long jobs restartable
Write state atomically, record what was finished, and make every step safe to run twice.