this article uses censored and truncated statistical models to assess technical effects potentially affecting the production frontier of embrapa's (empresa brasileira de pesquisa agropecuária) research centers. specifically the article assesses the significance of the factors level of research funds generated externally to the national treasury, actions of partnership, technical quality of research projects, client satisfaction, change in administration, type and size on the measurement of technical efficiency of the research centers. efficiency measures are computed both using data envelopment analysis (dea) and stochastic frontier methods. it is concluded that the dea efficiency measurements provide more information regarding the assessment of technical effects and that revenue generation, actions of partnership, technical quality of projects, and size are significant factors. the statistical analysis leading to these results fits a tobit model, defined by a distribution in the gamma family, to dea technical efficiency measurements. this analysis is consistent with the fit of a truncated normal distribution to residuals obtained from the fit of a dea production function.