convex text edit

[ajdamico]

[8/15, 11:56 AM] Anthony Damico: Brzezinski (2014Brzezinski, Michal. 2014. “Statistical Inference for Richness Measures.” Applied Economics 46 (14): 1599–1608. https://doi.org/10.1080/00036846.2014.880106.) also studied the reliability of the inference based on these estimators using a model-based Monte Carlo simulation, which are also valid for (design-based) simple random sampling with replacement. His results showed that inference for convex measures are unreliable for convex measures — see vignette below for a design-based approach with the same results.

i'm confused by that bold text -- is there a slightly different way to say that?

[guilhermejacob]

"... also studied the reliability of the inference based on these estimators using a model-based Monte Carlo simulation, which are also valid for (design-based) simple random sampling with replacement. His results showed that inferences for convex measures are unreliable — see vignette below for a design-based approach with the same results."

It means that you generally have an estimator that produces bad results. It is highly sensitive to outliers and confidence intervals are invalid (it says 95%, but actual coverage is much smaller). The simulation is also useful for explaining these problems, I believe.

[ajdamico]

@guilhermejacob that edit OK?

[guilhermejacob]

Yes. Can you update it in the book?