Measuring latent causal structure

Ricardo Silva

in Causality in the Sciences

Published in print March 2011 | ISBN: 9780199574131
Published online September 2011 | e-ISBN: 9780191728921 | DOI:
                      Measuring latent causal structure

Show Summary Details


The presence of latent variables makes the task of estimating causal effects difficult. In particular, it might not even be possible to record important variables without measurement error, a common fact in fields such as psychology and social sciences. A fair amount of theory is often used to design instruments to indirectly measure such latent variables, such that one obtains estimates of measurement error. If the measurement error is known, then causal effects can be identified in a variety of scenarios. Unfortunately, a strictly theoretical approach for formalizing a measurement model is error prone and does not provide alternative models that could equally or better explain the data. The chapter introduces an algorithmic approach that, given a set of observed indicators of latent phenomena of interest and common assumptions about the causal structure of the world, provides a set of measurement models compatible with the observed data. This approach extends previous results in the literature which would select an observed variable only if it measured a single latent variable. The extensions cover cases where some variables are allowed to be indicators of more than one hidden common cause.

Keywords: latent variable modeling; measurement error; causal discovery; structural equation models

Chapter.  10095 words.  Illustrated.

Subjects: Logic

Full text: subscription required

How to subscribe Recommend to my Librarian

Buy this work at Oxford University Press »

Users without a subscription are not able to see the full content. Please, subscribe or login to access all content.