Table 3

Comparison of Models M1-M3

model

age groups

DIC

M1 (binomial)

1

0.50

2870.2

M1 (binomial)

5

0.55, 0.52, 0.49, 0.53, 0.41

2872.3

model

Age groups

var(q)

DIC

M2 (beta-bin.)

1

0.51

0.003

2871.5

M2 (beta-bin.)

5

0.51, 0.46, 0.46, 0.47, 0.46

0.004,0.005,0.005,0.005,0.005

2871.3

Model

age groups

q₀

q₁

DIC

M3 (Markov)

1

0.50

0.47

0.59

2856.4

M3 (Markov)

5

0.55, 0.53,0.49, 0.53, 0.41

0.45, 0.44, 0.47, 0.56, 0.46

0.65, 0.64, 0.56, 0.55, 0.45

2860.0

Model results: M3 without age groups fitted the data best, indicated by it's lowest value of DIC. M2 effectively reduced to M1, as it estimated very low variance of detectability. All models estimated a true number of clones of approx. 546 (not shown), and similar values for the mean detectability Values of represent the bias-corrected mean detectability, whereas q₀ and q₁ are not bias corrected, but represent "observed" detectabilities (see section "bias correction").

Bretscher et al. Malaria Journal 2010 9:234   doi:10.1186/1475-2875-9-234