explanatory variables in a model provides a guide to the amount of selection on the unobservables. We also propose an informal way to assess selectivity bias based on measuring the ratio of selection on unobservables to selection on observables that would be required if one is to attribute the entire effect

Start with a probability measure e on a space of observable (“empirical”) variables. Can we find (i) an extended space that includes “hidden” variables, and (ii) a probability measure p on this space, where p is required to satisfy certain independence conditions, so that e can be obtained from p

Abstract. Bayesian networks (BNs) are an appealing model for causal and non-causal dependencies among a set of variables. Learning BNs from observational data is challenging due to the nonidentifiability of the network structure and model misspecification in the presence of unobserved (latent

Current psychological models of causal induction assume that causal relationships are inferred based on observations about whether the cause and effect are present or absent. The current study investigated how people infer the causal roles of unobserved events. In Experiment 1 we demonstrate

Abstract Unobserved variables correlated with market share are largely responsible for the high profitability of market share leaders; yet, very little is known about these unobserved variables. The objective of this paper is to empirically determine the cost/sales ratios through which unobserved

We present methodology for identifying the most likely configuration of unobserved variables, given observed variables, for collections of discrete and continuous variables with conditional Gaussian (CG) distributions. The dependency structure on the joint distribution of discrete and continuous

also propose using recently-introduced internet search volume indexes as proxies for fundamental locational traits, and discuss their advantages and limitations. We then successfully proxy for a number of economic and demographic variables in US cities and states. We further obtain document

-sectional variation in the occurrence-frequency of social phenomena. We characterize the conditions under which such proxying is likely. We successfully proxy for a number of demographic variables at the US city and state levels. We obtain document-frequencybased measures of corruption at the country and state level

We develop new estimation methods for the causal effect of a variable based on the idea that the amount of selection on the observed explanatory variables in a model provides a guide to the amount of selection on the unobservables. Our approach involves the use of factor model as a way to infer

. The EM algorithm performs maximum likelihood estimation for data in which some variables are unobserved. We present a function that resembles negative free energy and show that the M step maximizes this function with respect to the model parameters and the E step maximizes it with respect