The goal of mreg is to implements the techniques of exact likelihood when the discrete outcome can be missing in a regression model. It is the accompanying software to the paper Bond S, Farewell V, 2006, Exact Likelihood Estimation for a Negative Binomial Regression Model with Missing Outcomes, Biometrics
You can install the released version of mreg from CRAN with:
And the development version from GitHub with:
This is a basic example which shows you how to solve a common problem:
library(mreg)
mod1 <- mreg( damaged~offset(log(intervisit.time))+esr.init,
data=public,patid=ptno,print.level=1, iterlim=1000 )
#> iteration = 0
#> Step:
#> [1] 0 0 0
#> Parameter:
#> [1] 0 0 0
#> Function Value
#> [1] 3902.458
#> Gradient:
#> [1] 3066.5542 690.1451 22153.0783
#>
#> iteration = 34
#> Parameter:
#> [1] -0.981767860 -6.696169800 0.005720807
#> Function Value
#> [1] 555.7446
#> Gradient:
#> [1] -4.469371e-05 -2.064622e-05 -2.745820e-03
#>
#> Last global step failed to locate a point lower than x.
#> Either x is an approximate local minimum of the function,
#> the function is too non-linear for this algorithm,
#> or steptol is too large.
mod1
#>
#> Call:
#> mreg(formula = damaged ~ offset(log(intervisit.time)) + esr.init, data = public, patid = ptno, iterlim = 1000, print.level = 1)
#>
#> Coefficients:
#> (Intercept) esr.init
#> -6.696170 0.005721
summary(mod1)
#>
#> Call:
#> mreg(formula = damaged ~ offset(log(intervisit.time)) + esr.init,
#> data = public, patid = ptno, iterlim = 1000, print.level = 1)
#>
#>
#> Coefficients:
#> Estimate S.E. Z-value Pr(>|Z|)
#> (Intercept) -6.696170 0.134108 -49.931 < 2e-16 ***
#> esr.init 0.005721 0.003218 1.778 0.0755 .
#> log.disp -0.981768 0.146987 -6.679 2.4e-11 ***
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
#>
#> Deviance: 1111.489