Figure 10

The TKF92 model [9], presented as a Hidden Markov Model. λ is the insertion rate, μ is the deletion rate, r is the parameter of the geometric distribution of inserted and deleted fragments, $\frac{1-e^{(\lambda -\mu )t}}{\mu -\lambda e^{(\lambda -\mu )t}}$. Emission probabilities are given by the time-continuous substitution model of the TKF92 model, the probability of joint emission of two characters equals the joint observation probability of two characters under the substitution model, single emissions follow the equilibrium distribution of the substitution process.