Concentration models

Let x denote a random variable from a lognormal distribution. Then, the log transformed variable y=ln(x) is normally distributed with μ and variance σ. The probability density function (p.d.f.) of y may be expressed as:

fy(y,p0,μy,σy2)=p0I(y;0)+(1p0)(1I(y;0))12πσyexp(yμy)22σy2

where p0=Pr(y<log(Xlor)),xlor is the limit of reporting and I(y;0) is an indicator function for y<log(Xlor). For p0=0 the p.d.f. of y reduces to the usual lognormal density. The left truncated density for ylog(Xlor) may be expressed as:

fy(y;μy,σy2)=12πσyexp(yμy)22σy2/(1Φ(z))

with Φ() the standard normal c.d.f. and z=(log(xlor)μy)/σz. Model parameters are estimated using maximum likelihood estimation based on the loglikelihood functions specified below. The loglikelihood functions are evaluated in R, using the optim algorithm to find estimates for μy,σy2 and p0.

Mixture zero spike and censored lognormal

The loglikelihood may be expressed as:

logL(p0,μy,σy2)=i=1n0log(p0+(1p0)Φ(zi))+n1log(1p02πσy)i=n0+1n(yiμy)22σy2

where yi=log(xi), Φ() is the standard normal c.d.f., z=(log(xi,lor)μy)/σy, zlor=(log(lor)μy)/σy with n0 number of censored values (xi<xi,lor),n1 number of uncensored values (xixi,lor) and xi,i=1n.

Multiple values for LOR are allowed.

Censored lognormal

When p0=0 the loglikelihood reduces to:

logL(μy,σy2)=i=1n0log(Φ(z))+n1log(12πσy)i=n0+1n(yiμy)22σy2

Multiple values for LOR are allowed.

Mixture censored spike and truncated lognormal

Ignoring the n0 values below xlor, the loglikelihood may be expressed as:

logL(μy,σy2)=n1log(1Φ(z))+n1log(12πσy)i=n0+1n(yiμy)22σy2

Only one value for LOR is allowed.

Mixture censored spike and lognormal

Ignoring the n0 values below xlor, the loglikelihood may be expressed as:

logL(μy,σy2)=n1log(12πσy)i=n0+1n(yiμy)22σy2

Only one value for LOR is allowed.