1
$\begingroup$

I have expenditure data in several regions, and for each of them i know mean expenditure, standard deviation and skewness in original scale. Since data are skewed i want to compute probability of being below certain expenditure level (for example 1500$) with log normal distribution. For example in one region i have mean expenditure m=2000$, sd=1000, and skweness=1.1. in that case can i use

plnorm(1500,2000,1000)

or i need to transform evertything in log scale first? In plnorm parameters are described "mean and standard deviation of the distribution on the log scale with default values of 0 and 1 respectively." If i need to transform it first, is the following correct:

m=2000
s=1000
lsm=log(m)-(1/2)*log((s/m)^2+1)
lssd=sqrt(log((s/m)^2+1))
plnorm(log(1500),lsm,lssd)

I guess this is very related to post How to calculate log-normal parameters using the mean and std of the given distribution but in my case i also need to be sure i am using plnorm/pnorm correctly.

$\endgroup$
1
$\begingroup$

You need to transform first.

You can check with

dat <- rlnorm(10^6, lsm, lssd) 
mean(dat) 
sd(dat)

which should give you values near $2000$ and $1000$

You can then use either

plnorm(1500, lsm, lssd)

or

pnorm(log(1500), lsm, lssd)

to give about $0.35$ (your version of plnorm(log(1500),lsm,lssd) would be wrong)

I think the skewness could be rather higher, perhaps about $1.625$ rather than the question's $1.1$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.