TODO

GET RID OF THOSE WARNINGS

validate_inputs function before calling pexp, dexp to evaluate likelihood








#  flexsurvmix

bring pfinal back or delete tests 





Handle left truncation and interval censoring at the same time? 

is logliki component of flexsurvreg objects the same as deviance residuals given by survreg?

Ongoing problems with Hessian in Weibull models (see p metcalfe example)

predictions


SURVREG

does 1/2 Surv indicator work, and TRUE/FALSE?  if so is n events reported correctly



OPTIONAL

Time dependent covs in summary.flexsurvreg

Tidy up summary functions, including bootstrap, covariate handling.

D can be non pos-def in QP (github #9). Worked around for now, but why does this happen?

d/hsurvspline at x=0, currently zero, should it be limit as x reduces to 0?

Warn about survival times of zero (or stop=start) which may be invalid for particular models and parameter values, but not always

make plot/summary do baseline values of factor covs and mean of continuous

Residuals, e.g. Cox-Snell

abstract newdata thing - default to baseline

Delta method for survival - a lot of distribution-specific faff.

inits and fixedpars named

anova / LR test method, like anova.survreg, anova.survreglist, using
   stats::stat.anova
could loglik component simply be used for these

cluster standard errors as in survreg

plot two groups for a numeric 0/1 variable

hHsurvspline: vectorise as dp

intervals in Surv are open on left, closed on right

other types of status: mstate

CIs as grey bands not lines

#are these buggy in survreg::Surv?
#time <- ifelse(time == Inf, NA, time)
#time2 <- ifelse(time2 == -Inf, NA, time2)

Newton-Raphson (analytic second derivs) available in maxLik::maxNR

allow time-dependent covs to influence the hazard in pmatrix.fs

any role for pseudo-observations?  (residual plots? regressing functions of pars directly on covs?)
http://smm.sagepub.com/content/19/1/71.full.pdf+html

piecewise exponential distribution

compare with rstpm2 package

Robust estimates of variance matrix?