Most of the important agronomic traits in crops, such as yield and quality, are complex traits affected by multiple genes with gene × gene interaction as well as gene × environment interaction. Understanding the genetic architecture of complex traits is a long-term task for quantitative geneticists and plant breeders who wish to design efficient breeding programs. Conventionally, the genetic properties of traits can be revealed by partitioning the total variation into variation components caused by specific genetic effects. With recent advances in molecular genotyping and high-throughput technology, the unraveling of the genetic architecture of complex traits by analyzing quantitative trait locus (QTL) has become possible. The improvement of complex traits has also been achieved by pyramiding individual QTL. In this review, we describe some statistical methods for QTL mapping that can be used to analyze QTL × QTL interaction and QTL × environment interaction, and discuss their applications in crop breeding for complex traits. The most important economic traits in crops are quantitative in nature. The genetic variation of quantitative traits is usually controlled by minor-effect genes and the environments as well as by epistatic effects between different genes and by interactions between genes and environments. Because of their complicated genetic architecture, quantitative traits are usually referred to as complex trait. The genetic architecture includes knowledge about the numbers and positions of the genes in the chromosomes, the magnitude of their effects, and the contributions of additive, dominance and epistatic effects [1].Most plant breeders and quantitative geneticists have long been interested in uncovering the genetic architecture of yield and other complex traits [2]. Understanding genetic architecture can provide insights into how the architecture is translated into genetic variation and selection response [3]. Based on diallele cross experiments, the genetic architecture of a trait has been analyzed by decomposing the trait variation into additive, dominance, and epistasis variances and studying their interaction variances with the environments [4][5][6][7]. The total genetic covariance between two traits can also be dissected into genetic covariate components. With the increasing availability of molecular markers, the quantitative genetics theory of complex traits and marker-assisted selection has undergone major advances. It is now possible to connect marker variations with phenotypic variations, and to dissect genetic architecture of traits into individual genetic variants. The selection of traits can also be achieved using individual quantitative trait loci (QTLs). In this review, we discuss some of the statistical methods used for QTL mapping of experimental populations and examine their applications in marker-assisted selection (MAS) for improving complex traits.