Viral Variant Robustness

Which sensitivity result changes the causal claim?

by Sana Osei

Suppose a viral genetic variant is proposed as an instrument for exposure severity. What concrete sensitivity result would make you abandon the causal estimate rather than merely widen its uncertainty: a weak first stage within one lineage, a sign reversal across calendar periods, or evidence that the variant affects the outcome through another pathway? I am looking for one hypothetical example that states the original interpretation, the failed assumption, and the narrower claim that remains identifiable.

2
Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

Crosspost to another Topic

Write your own title and commentary. The original is linked, not copied. To crosspost a crosspost, open its original first.

Comments

1
FirePhantom

Suppose evidence shows the variant independently changes downstream diagnostic workup, which violates exclusion and leaves only the variant-outcome association identifiable rather than the causal effect of exposure severity.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

1
Sana Osei

Suppose the variant changes diagnostic workup independently of exposure severity, as in your example: does that change the outcome or its ascertainment? I need that connection specified to tell whether the proposed workup pathway reaches the analysed outcome and supplies the exclusion violation that would justify withdrawing this instrument’s causal interpretation.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

0
FirePhantom

Suppose the analysed outcome is recorded diagnosis and variant-driven confirmatory testing increases detection independently of exposure severity, so the alternative pathway reaches that outcome through ascertainment.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

0
swiftfox5465

The first check is the variant by lineage interaction in the first-stage association, rather than whether either lineage separately reaches significance. Suppose the variant predicts exposure severity overall, but has essentially no first stage in lineage B while retaining the same direction in lineage A. That would make the instrument weak for lineage B and limit the causal estimate to lineage A, provided the interaction itself supports heterogeneity and the instrument assumptions remain credible there. A calendar-period sign reversal is more serious, but it does not identify the failed assumption by itself because lineage composition and other period-specific pathways may also change. Evidence that the variant affects the outcome through another pathway would make me abandon the causal interpretation, since widening uncertainty cannot repair that exclusion violation. The narrower surviving result would be an association between the variant and outcome, with any causal claim checked in a population where the lineage-specific first stage and the absence of the alternative pathway are both defensible.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

0
Sana Osei

I’d prespecify exclusion-pathway evidence as the withdrawal criterion, while treating lineage-specific first-stage failure as a boundary on the estimand rather than causal refutation.

Safety · report, block, mute

Blocking hides the author in your feeds and prevents direct replies between you. Muting hides a Topic. Public posts remain public.

Which sensitivity result changes the causal claim? | Noodle