Coherent Identifier About this item: 20.500.12592/5bfc6r

A Count Data Model with Social Interactions




When the distribution of the outcome is almost degenerated such that the outcome takes only two values, the structure of the game and the BNE are similar to Lee et al. [...] Moreover, I provide a way to properly estimate the variance of the estimator of the outcome model, which takes into account the uncertainty of the estimation at the first stage.5 The remainder of the paper is organized as follows. [...] Therefore, I expect that estimating a SART model with data generated following the count response model would involve classical bias: the smaller the dispersion of the dependent variable, the greater the bias. [...] The results can be replicated using my R package CDatanet and the replication code.12 The Monte Carlo results show that the NPL estimator of the count data model performs well in finite samples regardless of the type of DGP (see Tables 2 and 3). [...] However, the bias of the SAR model is still large for the DGP of type A. Indeed, the SAR model does not control for the left-censoring nature of the dependent variable.


economics science and technology mathematics social sciences statistics standard deviation standard errors errors and residuals computing and information technology parameter estimator sd game statistical theory variance likelihood function ?xed e?ects overdispersion monte carlo method autoregressive negative binomial tobit autoregressive models underdispersion