Most physical traits of agricultural and medical importance are complex, meaning that they are determined by multiple genetic and environmental factors. For decades it has a been a major goal in biology to be able to understand and predict complex phenotypes based on an individuals genomic sequence. Modern genotyping technologies have enabled the collection of massive samples of paired genotype-phenotype data. Despite this deluge of data, the genetic basis of complex traits remains unclear. Here I attempt to address this problem through a detailed simulation study based on explicit population genetic models for the maintenance of heritable phenotypic variation for a complex disease trait. The main conclusion of this study is that gene-based recessivity, under which compound heterozygotes can have excess disease risk, should be a leading candidate to explain some otherwise perplexing statistical properties of complex diseases. I complement this simulation study with an implementation of a statistical method found to be more powerful under a gene-based recessive genetic architecture. I then further support the assumed fitness model utilized in the simulation study through an empirical analysis of selection in a contemporary human population. Through studying the relationships between phenotypes and lifetime reproductive success, I showed that weak stabilizing selection is common on human traits and that many traits of clinical significance are under directional selection.