About the Society
Papers, Posters, Syllabi
Submit an Item
Polmeth Mailing List
Below results based on the criteria 'unobservability'
Total number of records returned: 2
The Split Population Logit (SPopLogit): Modeling Measurement Bias in Binary Data
limited dependent variables
This study describes a split population logit model that can be useful to researchers who are modeling a binary dependent variable that is measured with a biased instrument. To motivate the study we identify two common, yet widely unrecognized, circumstances in which political scientists are likely to study dichotomous variables that have been measured with bias. In one such setting (e.g., surveys) the strategic interests of actors will lead them to misrepresent an attitude or behavior. In another such setting (e.g., content analysis of events) researchers' instruments are unable to distinguish between the absence of a characteristic or event and missing data. We briefly argue that "unobservability," "zero-inflated," and other models form a single class of models that allow researchers to model the bias in operational instruments, and thus not only correct bias in statistical inference but, more importantly, produce theoretical accounts of the bias and then test the hypotheses that those accounts imply. We derive the likelihood function for the split population logit model, describe the properties of its MLEs, present the results from a Monte Carlo study, and briefly describe code that researchers can use to implement the model in the Stata statistical package.
A Theoretical and Statistical Model of Indirect (Unobserved) Network Relationships
The study of political science has derived great benefits from the recent growth of conceptualizing and modeling political phenomena within their broader network contexts. More than just a novel approach to evaluating the old puzzles, network analysis provides a new way of theoretical thinking. In contrast to traditional treatment of non-independence among observations as a ``nuisance'', network theories view actors' positional or relational dependencies as the primary focus of analysis. Fast-paced progress in statistical modeling of networks has not been matched, however, by equal advances in theoretical understanding of many types of network outcomes, especially higher-order (indirect) network relationships (e.g., triads, 2-stars, 4-cycles). Despite the increasing ability to statistically model higher-order network complexities, the causal and theoretical processes associated with these complexities are poorly understood. This paper takes a first step towards a richer theoretical understanding of such complexities, by zeroing in on the causal processes for formation of indirect ties between nodes. Indirect ties may form as mere artifacts of a network (i.e., if there is a tie between actors A and B and actors B and C, then there is, by construction, an indirect tie between A and C), or as a purposeful channel for inter-mediated interaction (i.e., actors A and C use B as an ``intermediary'' in their interaction). Simply using a measure, such as triads, would conflate these two theoretically distinct processes (the latter is strategic, while the former is ``accidental'') and add little to our substantive understanding. Moreover, our ability to isolate the effects of the ``intermediated'' interactions in this example would be hindered by the presence of a (large) number of ``accidental'' indirect ties. I explore the two types of indirect ties on the example of the network of trade among international states and propose a statistical estimator that probabilistically separates the types of indirect ties using two sets of exogenous covariates. Finally, I evaluate the properties of the proposed estimator using Monte Carlo simulations.