Instrumental variable methods in structural equation models

Instrumental variable regression (RegIV) provides a means for detecting and correcting parameter bias in causal models. Widely used economics, recently several papers have highlighted its potential utility ecological applications. Little attention has thus far been paid to the fact that IV methods can also be implemented within structural equation models (SEMIV). In this paper I present motivations, requirements basic procedures using SEMIV. first consider inference IVs from perspective of randomized experiment with partial control cause interest. common sources bias, role randomization limits capacity exclude bias. Sources include omitted confounders, reciprocal causation, reverse causation measurement error, all which seen as single problem—endogeneity. The approach estimating most commonly econometric practice, two-stage least squares (2SLS), is explained, followed by brief exposition covariance modelling SEM. Using data an field experiment, illustrate use treatment then evaluating candidate variables might serve additional IVs. are shown useful both endogeneity removing influences. some ways generated, well diagnostic capabilities remedy embedded Procedures screening reveal valuable lessons regarding theoretical empirical standards SEMIV way detect suggest SEM framework support simultaneous pursuit explanatory modelling, pair aspirations ecologists. Moving forward, there need better understanding successful application.