This is a perspective similar to DanielLC’s point. Additionally, a commenter there makes the parallel point that we don’t really know whether private insurance improves the outcome measures.
Your argument is one that was only made after the result was known; this is a classic failure mode.
True, but we shouldn’t overstate the argument. The p-values were not low enough to count as “statistically significant,” but the direction of change was towards improved health outcomes. One is doing something wrong with this evidence if one updates against improved health outcomes for public health insurance for the poor (i.e. Medicaid).
One is doing something wrong with this evidence if one updates against improved health outcomes for public health insurance for the poor (i.e. Medicaid).
Updates always move you towards what you just saw, and so if your estimate was above what you just saw, you update down. If you only consider the hypotheses that Medicaid “improves,” “has no effect,” or “harms,” then this is weak evidence for “improves” (and “has no effect”). But a more sophisticated set of hypotheses is the quantitative effect of Medicaid; if one estimated beforehand that Medicaid doubled lifespans (to use an exaggerated example), they should revise their estimate downward after seeing this study.
Fair enough. I should have said “McArdle and her political allies are making a mistake by not updating towards ‘Medicaid improves health outcomes,’” given my perception of their priors.
This is a perspective similar to DanielLC’s point. Additionally, a commenter there makes the parallel point that we don’t really know whether private insurance improves the outcome measures.
True, but we shouldn’t overstate the argument. The p-values were not low enough to count as “statistically significant,” but the direction of change was towards improved health outcomes. One is doing something wrong with this evidence if one updates against improved health outcomes for public health insurance for the poor (i.e. Medicaid).
Updates always move you towards what you just saw, and so if your estimate was above what you just saw, you update down. If you only consider the hypotheses that Medicaid “improves,” “has no effect,” or “harms,” then this is weak evidence for “improves” (and “has no effect”). But a more sophisticated set of hypotheses is the quantitative effect of Medicaid; if one estimated beforehand that Medicaid doubled lifespans (to use an exaggerated example), they should revise their estimate downward after seeing this study.
Fair enough. I should have said “McArdle and her political allies are making a mistake by not updating towards ‘Medicaid improves health outcomes,’” given my perception of their priors.