←back to thread

231 points cachecrab | 1 comments | | HN request time: 0.208s | source
Show context
srvmshr ◴[] No.31900386[source]
At some point we (researchers & scientists) should draw the line between correlation & causation.

That there is 40% less incidence in flu vaccinated patients is established by observation, but is it due to flu vaccine by itself or factors which surrounds itself in people more prone to flu or who are medically more conscious? (edit: think of analysis of variance tests among identifiable medical & socioeconomic variables). If the researchers establish a positive link, wouldn't it be worthwhile to then suppose that amyloid plaques are somehow related to viral influenza replication - which by itself sounds a bit absurd (edit: Ok lot of good reference about infections around AD. I take this back, my bad)

The alternative hypothesis is worth thinking too: Are people more prone to Alzheimer's also more prone to catching flu.

replies(2): >>31900434 #>>31900495 #
lukev ◴[] No.31900495[source]
At some point we (commenters on hacker news) should stop reflexively typing this in response to every study without reading the paper.

From the abstract:

> Propensity-score matching (PSM) was used to create flu-vaccinated and flu-unvaccinated cohorts with similar baseline demographics, medication usage, and comorbidities.

If you want to argue that the authors did a bad job at this, for some reason, that would be a really meaningful contribution and I'd love to hear the details.

Otherwise, can we assume that the scientists writing and reviewing papers did a semi-competent job and are at least as good at analyzing quantitative claims as someone commenting on HN because they are bored at work?

replies(2): >>31900736 #>>31900819 #
1. srvmshr ◴[] No.31900736[source]
I am not an expert at this, though I have taken few credits of Biostatistics. I would be more interested if swapping the variables also gave significance in tests. Did the Alzheimer's cohort also have a higher incidence of flu related comorbidities? I understand this is difficult - but this is why ANOVA tests also are extensively used to show variable dependence or report the degree of dependence. It is unclear from their findings. Without that we are probably looking at some positive correlation without investigating the causation.