How to estimate Location and Scale of lognormal distribution using Survreg I have created a lognormal survival model (via survreg in the survival package in R). 
How can I estimate the location and scale parameters of a lognormal survival model like this directly (i.e. without using something automatic like the predict function)?
I am trying to use the location and scale parameters to calculate the expected value using the method of moments.
hmohiv<-read.table("http://www.ats.ucla.edu/stat/r/examples/asa/hmohiv.csv", sep=",", header = TRUE)

library(survival)
test <- survreg( Surv(time, censor) ~ age, dist="lognormal")
summary(test)

Here is the output:
> summary(test)

Call:
survreg(formula = Surv(time, censor) ~ age, dist = "lognormal")
              Value Std. Error     z        p
(Intercept)  5.0844     0.6531  7.79 6.95e-15
age         -0.0873     0.0178 -4.92 8.74e-07
Log(scale)   0.1260     0.0793  1.59 1.12e-01

Scale= 1.13 

Log Normal distribution
Loglik(model)= -271.1   Loglik(intercept only)= -281.9
        Chisq= 21.47 on 1 degrees of freedom, p= 3.6e-06 
Number of Newton-Raphson Iterations: 4 
n= 100 

 A: I originally anticipated this was an R question (and so the request for a reproducible example was to make it migratable to stackoverflow), but now that you've clarified a little I see there's a statistical issue first and foremost.
I will explain the underlying statistical issues, which are on topic here.
You are not fitting a single lognormal distribution, but a collection of them -- a different one to every point.
Let's start with a much simpler case: imagine you were to fit a normal regression model $y_i=\beta_0+\beta_1 x_i + \varepsilon_i$, where the $\varepsilon_i$'s are iid $N(0,\sigma^2)$.
Then $Y_i|x_i \sim N(\beta_0+\beta_1 x_i,\sigma^2)$
That is, the scale parameter is the same for every observation, but the location differs. The estimated location and scale parameters for each observation are then obtained by replacing those population parameters by their estimates.
It's exactly the same here, but now you're modelling the log of the survival time as a conditionally normal r.v. (and here your fitting takes account of the censoring).
The location parameters of the normal distributions for log(time) -- and hence the location parameters of the lognormal -- should be given by $\beta_0+\beta_1 x_i$ where $x_i$ is the age of the $i$th person and the coefficients are exactly the ones that appear in the output. Similarly, to my understanding, the estimated scale parameter is that given as "Log(scale)" which if I understand correctly is not the log of the estimated scale parameter but the estimated scale parameter of the log(time) distribution (i.e. I think that's $\hat{\sigma}$ in the output).
The calculation, then, of the $n$ location parameters is simply a matter of substituting in the formula.
