The congruence–bias parameterisation
Source:vignettes/articles/reparameterisation.Rmd
reparameterisation.RmdThe widget below demonstrates the recommended congruence–bias
parameterisation by simulating ten pairs of counts from a model (see
rbicount()). The relationship with the native
parameterisation is shown by moving the sliders: a change to a parameter
in one parameterisation changes the parameters in the other. The plot
below the sliders shows the expected decomposition of the two counts
(left) and the simulated pairs (right).
Congruence–bias parameterisation
Native parameterisation
The coordinates of the congruence–bias parameterisation directly represent the quantities usually of interest in a comparison of two sources:
| Coordinate | Meaning | Definition | Name in brms |
|---|---|---|---|
| M | midpoint of the expected counts of the two sources | M = \mu + (\lambda_1 + \lambda_2)/2 |
logM, for \log M
|
| f | congruence, the proportion of M recorded by both sources | f = \mu / M |
logitf, for \operatorname{logit} f
|
| \beta | source bias, zero when the two exclusive rates are equal | \beta = (\lambda_1 - \lambda_2)/(\lambda_1 + \lambda_2) |
delta, for \operatorname{artanh}\beta
|
| \kappa_s, \kappa_{x1}, \kappa_{x2} | overdispersions | \kappa = 1/\sqrt{\phi} |
kappas, kappaxone,
kappaxtwo
|
The native parameters are the rates and dispersions of the three latent components:
| Parameter | Meaning | Definition | Name in brms |
|---|---|---|---|
| \mu | rate of the shared component, recorded by both sources | \mu = Mf | mu |
| \lambda_1 | rate of the source-exclusive component of source 1 | \lambda_1 = M(1 - f)(1 + \beta) | lambdaone |
| \lambda_2 | rate of the source-exclusive component of source 2 | \lambda_2 = M(1 - f)(1 - \beta) | lambdatwo |
| \phi_s, \phi_{x1}, \phi_{x2} | NB2 (inverse) dispersions | \phi = 1/\kappa^2 |
shapes, shapexone,
shapextwo
|
The functions mfd_to_dpars() and
dpars_to_mfd() convert between the two parameterisations,
for single values or for vectors of posterior draws:
library(bicountbrms)
dpars <- mfd_to_dpars(
M = 12,
f = 0.67,
delta = atanh(0.2),
kappas = 0.6,
kappax = 1.0
)
unlist(dpars)
#> mu lambdaone lambdatwo shapes shapexone shapextwo
#> 8.040000 4.752000 3.168000 2.777778 1.000000 1.000000
mfd <- dpars_to_mfd(
mu = dpars$mu,
lambdaone = dpars$lambdaone,
lambdatwo = dpars$lambdatwo
)
unlist(mfd[c("M", "f", "beta")])
#> M f beta
#> 12.00 0.67 0.20