%% what I want to do:
x_true % "true", error-free measurement of a variable over time
x_noise % synthetic noise, not necessarily homoskedastic and/or normally distributed
x = x_true + x_noise
for i = 1:n_boot
    p_est(i) = parameter for best fit of model to x
    x = x_true + resample( x_noise ) % bootstrap x_noise (I guess wild bootstrap would be best here?)
end

%% what I believe "normal" bootstrapping of residuals would look like ???
for i = 1:n_boot
    p_est(i) = parameter for best fit of model to x
    x = x_first_best_fit + resample( x_first_best_fit - x )
end