Multilevel factor models: Identification of Three-level Model Parameters for the Study of Regional Development in Argentina

Palermo

Congreso; International Worshop in Statistical Modeling; 2013

International Worshop in Statistical Modeling

Abstract: This work is based on theoretical-social model which states that Communication and Information Technologies (CITs) infuence the human development directly and indirectly, due to the impact they have on economic growth. The aim of this work is to explore the relationship between the social, economic and technological dimensions or constructs, in Argentina. Confirmatory common framework (GLLAMMs, Skrondal & Rabe-Hesketh, 2004) was performed considering three level modeling. This approach allowed us to identify some indicators that measure both constructs and help characterize the level of development of our region using hierarchical information. Since the estimation method was based on full information maximum likelihood, we devoted special attention to the identifiability of the parameters, based mainly on Bollen (1989) rules. These rules apply to models with a factor complexity of one. Our model satisfies this condition in each level and, as it will be explained, this fact is sufficient because the process of identifiability is essentially hierarchical. The successive modeling steps led us to decide on significant indicators in the conformation of the constructs. Two constructs to describe the socioeconomic and technological (SET) development at the district level were obtained. Due to correlation between the two latent variables at department level was near one, a new common factor model containing only one dimension was appropriate.