/
SLsumm.R
152 lines (133 loc) · 6.97 KB
/
SLsumm.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
if(grepl('stevebe', Sys.info()['nodename'])) setwd('~/Documents/R Repos/EbolaVaccSim/')
if(grepl('stevebellan', Sys.info()['login'])) setwd('~/Documents/R Repos/EbolaVaccSim/')
if(grepl('tacc', Sys.info()['nodename'])) setwd('/home1/02413/sbellan/VaccEbola/')
library(RColorBrewer); library(boot)
## Simulate SWCT vs RCT vs CRCT for SL
sapply(c('simFuns.R','AnalysisFuns.R','CoxFxns.R','EndTrialFuns.R'), source)
thing <- 'SLSimsFinalPTCorr'
batchdirnm <- file.path('BigResults',thing)
fls <- list.files(batchdirnm, pattern='.Rdata', full.names = T)
fls <- fls[grepl('pit', fls)]
length(fls)
dparms <- c('trial','sdLogIndiv','vaccEff','doSL','propInTrial','nbsize','ord','reordLag','delayUnit','immunoDelay','trialStartDate'
, 'weeklyDecay', 'cvWeeklyDecay', 'cvClus', 'cvClusTime', 'remStartFin', 'remProtDel'
)
nbatch <- length(fls)
finInfoList <- finModList <- stopList <- parmsList <- list(NULL)
for(ii in 1:nbatch) {
if(ii%%100 ==0) print(ii)
ff <- fls[ii]
if(exists('sim')) rm(sim)
load(ff)
if(exists('sim')) {
sim$parms[['trialStartDate']] <- as.character(sim$parms[['trialStartDate']])
parmsList[[ii]] <- data.frame(nbatch = ii, t(unlist(sim$parms[dparms])))
tmpMod <- data.frame(nbatch = ii, sim$sim$finMods)
finModList[[ii]] <- merge(tmpMod, sim$sim$finInfo, by = 'sim')
}
}
parmsDT <- rbindlist(parmsList, use.names = T, fill = T)
finTrials <- merge(rbindlist(finModList), parmsDT, by = c('nbatch'))
finTrials[,vaccEff := levels(vaccEff)[vaccEff]]
finTrials[, sum(is.na(p)), mod]
finTrials[, sum(is.na(lci)), mod]
finTrials[, length(lci), list(propInTrial, mod)]
finTrials[mod=='coxME' & is.na(p), err:=1] ## sometimes cox returns NaNs, or partial NA's for certain values
finTrials$vaccEff <- as.numeric(finTrials$vaccEff)
## ## Simulations with less than 10 cases are considered to not have any power
## finTrials$tooSmall <- finTrials[, (caseCXimmGrpEnd + caseVXimmGrpEnd) < 10]
## finTrials[tooSmall==T, c('vaccGood','vaccBad','stopped') := F]
## finTrials[tooSmall==T, c('lci','uci','p') := list(-Inf,1,1)]
## Determine if stopped
finTrials[grepl('boot',mod), stopped := lci > 0 | uci < 0]
finTrials[grepl('relab',mod), stopped := p < .025]
finTrials[!grepl('boot',mod) & !grepl('relab',mod), stopped := p < .05]
finTrials[, vaccGood := stopped==T & mean > 0]
finTrials[, vaccBad := stopped==T & mean < 0]
## Coverage
finTrials[, cvr := lci < vaccEff & uci > vaccEff]
finTrials[is.na(cvr), cvr := F]
finTrials[grepl('relab',mod), cvr := NA] # no CI's for perm test
## Bias, must be done on RH/(RH+1) scale to deal with Inf & 0 RH's
finTrials$RH <- finTrials[, 1-mean]
finTrials$PHU <- finTrials[, RH/(RH+1)] ## proportion of hazard unavoidable even with vaccination
finTrials[RH==Inf, PHU:=1] ## otheriwse gives NaN for Inf/(Inf+1)
finTrials[,list(vaccEff,mean,PHU)] ## NEED TO AVERAGE BIAS ON PHU scale
## Reorder columns
front <- c('mod','stopped','vaccGood','vaccBad')
setcolorder(finTrials, c(front, setdiff(names(finTrials), front)))
back <- c('nbatch','sim')
setcolorder(finTrials, c(setdiff(names(finTrials), back), back))
save(finTrials, file=file.path('BigResults', paste0(thing, '.Rdata')))
load(file=file.path('BigResults',paste0(thing, '.Rdata')))
powFin <- summarise(group_by(finTrials, vaccEff, trial, propInTrial, ord, delayUnit, mod, immunoDelay,trialStartDate
, weeklyDecay, cvWeeklyDecay, cvClus, remProtDel, remStartFin) #, cvClusTime)
, nsim = length(stopped)
, stopped = mean(stopped)
, vaccGood = mean(vaccGood)
, cvr = mean(cvr)
, cvrNAR = mean(cvr, na.rm=T)
, mean = mean(mean)
, meanNAR = mean(mean, na.rm=T)
, vaccBad = mean(vaccBad)
, stoppedNAR = mean(stopped,na.rm=T)
, vaccGoodNAR = mean(vaccGood,na.rm=T)
, vaccBadNAR = mean(vaccBad,na.rm=T)
, PHUNAR = mean(PHU, na.rm=T)
, meanErr = mean(err)
, meanBump = mean(bump)
, stopDay = mean(stopDay)
, caseTot = mean(caseCXimmGrpEnd + caseVXimmGrpEnd )
, caseC = mean(caseCXimmGrpEnd)
, caseV = mean(caseVXimmGrpEnd)
)
powFin[,propInTrial:= as.numeric(levels(propInTrial)[propInTrial])]
powFin[,delayUnit:= as.numeric(levels(delayUnit)[delayUnit])]
powFin[,trial:=factor(trial)]
## Get bias from means done on a PHU scale
powFin$RHxPHUNAR <- powFin[, -PHUNAR/(PHUNAR-1)]
powFin$meanXPHUNAR <- powFin[, 1 - RHxPHUNAR]
powFin$biasNAR <- powFin[, meanXPHUNAR - vaccEff]
powFin[vaccEff==.7, list(meanNAR,meanXPHUNAR,vaccEff, biasNAR)]
## Formatting stuff
front <- c('mod','vaccEff','stoppedNAR','vaccGoodNAR','cvrNAR','biasNAR',
'nsim','meanErr','propInTrial','vaccBad','cvr','stopped','vaccGood')
setcolorder(powFin, c(front, setdiff(names(powFin), front)))
pf <- data.table(powFin)
pf <- pf[!(trial=='FRCT' & delayUnit==0) & !(ord=='TU' & delayUnit==0)] ## redundant
pf$trialStartDate <- as.Date(pf$trialStartDate)
pf[mod=='coxME', mod:='CoxME']
pf$mod <- factor(pf$mod, levels=unique(pf$mod))
pf$order <- pf$ord
pf[delayUnit==0, order:='simultaneous instant']
pf$design <- pf$trial
levels(pf$design)[levels(pf$design) == 'SWCT'] <- 'SWT'
levels(pf$order)[2] <- 'time-updated'
pf[, immunoDelay:=as.numeric(levels(immunoDelay)[immunoDelay])]
pf[, pit:=factor(paste0(propInTrial*100,'%'))]
pf[, pit:=factor(pit, levels = c('2.5%','5%','7.5%','10%'), ordered = T)]
baseMods <- c('Cox PH Frailty'
, 'Poisson GLM\n no cluster effects'
, 'Poisson GLM \nwith fixed effects by cluster')
pf$model <- pf$mod
levels(pf$model) <- paste0(rep(c('', 'bootstrap over\n', 'permutation test over\n'),each=3), rep(baseMods,3))
save(pf, file=file.path('Results',paste0('powFin_',thing,'.Rdata')))
## to delete a range of jobs
## qdel echo `seq -f "%.0f" 2282389 2282404`
## Combine results from several analysis for plotting
thing <- 'SLSimsFinal'
load(file=file.path('Results',paste0('powFin_',thing,'.Rdata')))
pfOld <- pf
pfOld$remStartFin <- pfOld$remProtDel <- as.logical(NA)
pfOld[trial=='SWCT', c('remStartFin','remProtDel') := F]
thing <- 'SLSimsFinalPTCorr' ## adding results with new SWCT pt calculations
load(file=file.path('Results',paste0('powFin_',thing,'.Rdata')))
pfNew <- pf
arrange(pfNew[immunoDelay==21, length(design), list(trial, remProtDel, remStartFin,propInTrial,immunoDelay)], propInTrial)
thing <- 'SWCTkeepStartFin' ## adding results with new SWCT pt calculations
load(file=file.path('Results',paste0('powFin_',thing,'.Rdata')))
pfSF <- pf
arrange(pfSF[immunoDelay==21, length(design), list(trial, remProtDel, remStartFin,propInTrial,immunoDelay)], propInTrial)
pf <- rbindlist(list(pfOld, pfNew, pfSF), use.names=T, fill=T)
arrange(pf[trial=='SWCT' & immunoDelay==21, length(mean), list(trial, remProtDel,remStartFin,propInTrial)], propInTrial)
save(pf, file=file.path('Results','powFin_All.Rdata'))