Abstract
Instrumental variable (IV) methods are used to estimate causal effects insettings with unobserved confounding, where we cannot directly experiment onthe treatment variable. Instruments are variables which only affect the outcomeindirectly via the treatment variable(s). Most IV applications focus onlow-dimensional treatments and crucially require at least as many instrumentsas treatments. This assumption is restrictive: in the natural sciences we oftenseek to infer causal effects of high-dimensional treatments (e.g., the effectof gene expressions or microbiota on health and disease), but can only run fewexperiments with a limited number of instruments (e.g., drugs or antibiotics).In such underspecified problems, the full treatment effect is not identifiablein a single experiment even in the linear case. We show that one can stillreliably recover the projection of the treatment effect onto the instrumentedsubspace and develop techniques to consistently combine such partial estimatesfrom different sets of instruments. We then leverage our combined estimators inan algorithm that iteratively proposes the most informative instruments at eachround of experimentation to maximize the overall information about the fullcausal effect.