In lectures/organization_capital.md, the cell producing the "Posterior mean convergence and uncertainty" figure simulates a single worker with T = 20. The text just below it says:
As the number of screening observations $n$ increases, the posterior mean converges to the true $\theta$, and the posterior uncertainty shrinks at rate $1/\sqrt{n}$.
The posterior mean in the left panel does not converge to θ_true = 0.8. It wanders and ends well away from it.
This is about the sample size rather than the seed: PR #1014 migrates the cell to np.random.default_rng(0), which changes the draws but not this behaviour. Screenshots of the figure before and after that migration are attached below.
Before the migration:
After the migration:
Two ways to make the figure match the text:
- Increase
T.
- Or soften the prose so that it describes what the figure shows at
T = 20.
CC @mmcky and @HumphreyYang.
In
lectures/organization_capital.md, the cell producing the "Posterior mean convergence and uncertainty" figure simulates a single worker withT = 20. The text just below it says:The posterior mean in the left panel does not converge to
θ_true = 0.8. It wanders and ends well away from it.This is about the sample size rather than the seed: PR #1014 migrates the cell to
np.random.default_rng(0), which changes the draws but not this behaviour. Screenshots of the figure before and after that migration are attached below.Before the migration:
After the migration:
Two ways to make the figure match the text:
T.T = 20.CC @mmcky and @HumphreyYang.