by Chris Fariss
Keith Schnakenberg and I continue work on a paper in which we develop a measure the unobservable level of respect for human rights. I blogged about it earlier here. In the new version of the paper we build upon existing insights in the Bayesian measurement literature to develop a dynamic ordinal item-response (DO-IRT) model. We then assess the validity of the estimates obtained from the latent variable generated from the DO-IRT model with the estimates from an ordinal item-response (O-IRT) model and the original additive human rights scales. Below is a plot of the estimates of the latent physical integrity variable (see the paper for a similar plot of latent empowerment variable):
Below is the current version of the abstract:
We develop a Bayesian statistical technique, the dynamic ordinal item response theory model, in order to estimate latent human rights variables from the two ordinal Cingranelli and Richards (CIRI) human rights indices (physical integrity rights and empowerment rights). The dynamic ordinal item-response model builds upon the ordinal item response theory model and dyanimic item response model. All of these models explicitly account for the measurement error that results when different component indicators are aggregated together. We present both models, which each produce point estimates and credible intervals for latent physical integrity respect and latent empowerment respect. We assess the validity of these latent variables in relation to each other and with existing human rights scales. We then present a simple method that allows measurement error to be included in the uncertainty of causal estimates and also demonstrate the utility of the new latent human rights variables by replicating a recent analysis that uses ordinal CIRI human rights variable as two of the main explanatory variables of interest. We conclude by recommending the inclusion of the new human rights scales in studies that posit a causal effect of human rights abuse on some outcome of interest.