Aiming to estimate the combining ability in tropical and temperate popcorn (Zea mays L. var. everta Sturt.) lines using multivariate analysis, ten popcorn lines were crossed in a complete diallel without reciprocals and the lines and hybrids were tested in two randomized complete block experiments with three replicates. Data were subjected to univariate and multivariate ANOVA, principal component analysis, and univariate and multivariate diallel analysis. For multivariate diallel analysis, variables were divided into group I (grain yield, mean weight of ears with grains, popping expansion, mean number of ears per plant, and final stand) and group II (days to silking, plant height, first ear height, and lodged or broken plants). The P2 line had positive values for agronomic traits related to yield and popping expansion for group I, whereas the P4 line had fewer days to silking and lodged or broken plants for group II. Regarding the hybrids, P2 x P7 exhibited favorable values for most of the analyzed variables and had potential for recommendation. The multivariate diallel analysis can be useful in popcorn genetic improvement programs, particularly when directed toward the best cross combinations, where the objective is to simultaneously obtain genetic gains in multiple traits.