Identifying relevant and irrelevant variables in sparse factor models

Abstract

This paper considers factor estimation from heterogeneous data, where some of the variables-the relevant ones-are informative for estimating the factors, and others-the irrelevant ones-are not. We estimate the factor model within a Bayesian framework, specifying a sparse prior distribution for the factor loadings. Based on identified posterior factor loading estimates, we provide alternative methods to identify relevant and irrelevant variables. Simulations show that both types of variables are identified quite accurately. Empirical estimates for a large multi-country GDP dataset and a disaggregated inflation dataset for the USA show that a considerable share of variables is irrelevant for factor estimation.

The technology acceptance scale: Its Bayesian psychometrics assessed in a factor analysis via Markov chain Monte Carlo models

“…In Scopus, altogether 33 articles cited this reference whereas only 20 of them stem from subjects that are related to HF/E: 9 are from social sciences (Camilli & Fox, 2015;J. Chen, 2021;Gielen et al, 2018;Hornungová & Milichovský, 2019;Kaufmann & Schumacher, 2017;Levy et al, 2020;Loehlin & Beaujean, 2016;Milichovský, 2015;Piatek & Pinger, 2016), 7 from engineering (Apte et al, 2020;Gielen et al, 2018;Hornungová & Milichovský, 2019;Hu & Yang, 2017;Milichovský, 2015;Valera et al, 2020;Wang et al, 2019) and corresponding in-depth data analysis. However, the subchapter explicitly demonstrates coding the procedure according to "BayesFM" in the R programming language (Mair, 2018, pp.…”

Bayesian nonparametric sparse VAR models

“…Following the literature on international business cycles we consider a multi-country macroeconomic dataset (e.g., see Kose et al, 2003;Francis et al, 2017;Kaufmann and Schumacher, 2017) and extract a network of linkages between the cycles of the OECD countries, by applying BNP-Lasso VAR.…”

“…In the last decade, high dimensional models and large datasets have increased their importance in economics (e.g., see Scott and Varian, 2014) and finance. In macroeconomics, some authors investigate the use of large datasets (between 10 and 170 series) to improve forecasts (e.g., see Banbura et al, 2010;Stock and Watson, 2012;Koop, 2013;Carriero et al, 2015;McCracken and Ng, 2016;Kaufmann and Schumacher, 2017), while in finance, large datasets (between 10 and 200 series) have been used to analyse financial crises, contagion effects and their impact on the real economy (e.g., see Brownlees and Engle, 2016;Barigozzi and Brownlees, 2018). 1 Moreover, the level of interaction, or connectedness, between financial institutions represents a powerful tool in monitoring financial stability (e.g., see Diebold and Yilmaz, 2015;Scott, 2016), whereas measuring interdependence in business cycles and in financial markets (Diebold and Yilmaz, 2014;Demirer et al, 2018) is essential in pursuing economic stability.…”

Forecast density combinations of dynamic models and data driven portfolio strategies

“…In all models, we follow the identification scheme in Lopes and West (2004) and assume diagonal covariance matrices. See Chan et al (2018) and Kaufmann and Schumacher (2017) for more recent specifications of identification in this class of models.…”

“…strategy are not explicitly included since the realized returns from this strategy were unstable for all models due to estimation uncertainty and potential ill-conditioning in variance-covariance matrix estimates, see also Michaud (1989). A fair inclusion of the results of this strategy requires more structured or 'sparse' variance-covariance matrix estimation as in Kaufmann and Schumacher (2017). This is left as a topic for further research.…”