TABLE 1

MATLAB simulations confirm analytic theory

SelectionAllelesPopulation sizeExpected ΦObserved ΦObserved f
Overdominant2100,0002.22.1981 (2.1 × 10−4)−0.0472 (1.1 × 10−5)
Underdominant2100,0001.81.8011 (1.5 × 10−4)0.0508 (2.1 × 10−5)
Neutral2100,00022.0005 (1.4 × 10−4)−0.0001 (8.9 × 10−6)
Stochastic fitness2100,000>22.0318 (6.9 × 10−2)−0.0037 (4.3 × 10−3)
Stochastic fitness3100,000>22.0010 (2.8 × 10−2)0.0013 (7.8 × 10−4)
Stochastic fitness4100,000<21.9898 (1.5 × 10−2)0.0016 (3.0 × 10−4)
Directional21,0002.09762.1639 (6.8 × 10−2)−0.0195 (8.9 × 10−4)
Directional210,0002.09762.0989 (6.3 × 10−3)−0.0149 (1.3 × 10−4)
Directional2100,0002.09762.1013 (6.2 × 10−4)−0.0156 (2.7 × 10−5)
Directional3100,0002.06462.0681 (1.3 × 10−3)−0.0115 (1.2 × 10−5)
Directional4100,0002.04822.0467 (2.6 × 10−3)−0.0097 (9.4 × 10−6)
  • Simulations were run for 100 generations and mean and variance of Φ were computed (with variance in observed Φ within parentheses). All alleles were equally frequent at the start of each simulation run. Fitnesses are as follows: overdominant selection (wij = 1.1, wii = 1.0), underdominant selection (wij = 0.9, wii = 1.0), stochastic fitness (fitnesses for each genotype were generated each generation from a Gaussian distribution with a mean of 1.0 and a standard deviation of 0.1), and directional selection (homozygotes and heterozygotes containing a favored dominant allele have a fitness of 1.1, while all other genotypes have a fitness of 1.0).