/
abcPlotPosterior.R
187 lines (170 loc) · 9.63 KB
/
abcPlotPosterior.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
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
####################################################################################################
## Plot posterior results of ABC for manuscript
## Steve Bellan, 2014, steve.bellan@gmail.com
####################################################################################################
rm(list=ls()); gc()
setwd('/home1/02413/sbellan/Rakai/AcuteRetroSim/')
sapply(c("SimulationFunctions.R","RakFunctions.R",'abcFunctions.R'), source) # load Rakai analysis simulation functions from script
ncores <- 12
finalbatch <- 5
transfEHM <- TRUE
parsDo <- c('acute.sc','dur.ac','EHMacute','bp','het.gen.sd')
logDo <- c(T,T,F,T,F) ## which variables to show on log scale
flatPriors <- c(T,T,F,T,T) ## which variables have flat priors
nmsDo <- expression( RH[acute], d[acute],EHM[acute],lambda, sigma[lambda] ) ## variable names
npars <- length(parsDo)
distrNms <- c('Prior', paste0('Intermediate ', 1:(finalbatch-1)), 'Posterior') ## distribution names
xlims <- list(acute.sc=c(.5,200), dur.ac=c(.5,10), EHMacute=c(-5,100,500),
bp = exp(c(logGHazLoBound['bp'], logGHazHiBound['bp'])),het.gen.sd=c(0,3))
xticks <- list(acute.sc=c(.2,1,5,10,50,100,200), dur.ac=c(.5,1:10), EHMacute=c(-5,1,2,5,10,20,50,100,200,500),
bp = signif(exp(seq(logGHazLoBound['bp'], logGHazHiBound['bp'], l = 5)),2), het.gen.sd=seq(0,3,by=.5))
flatDens <- list(rhAcutePrior(parms = 5), durAcutePrior(parms=3),EHMacute=NA,
bp = logGHazPrior(parms=logGHazPrior(1))[3], het.gen.sd = hetGenSDPrior(parms=2))
logxticks <- xticks
for(ii in which(logDo)) logxticks[[ii]] <- log(xticks[[ii]])
rgs <- list(acute.sc=c(.5,200), dur.ac=c(.5,8), EHMacute=c(-5,500), bp=exp(c(logGHazLoBound[3], logGHazHiBound[3])), het.gen.sd=c(0,3))
for(ii in which(logDo)) rgs[[ii]] <- log(rgs[[ii]])
if(transfEHM) {
rgs[['EHMacute']] <- log(rgs[['EHMacute']]+10)
ehmticks <- log(xticks[['EHMacute']]+10)
xlims[['EHMacute']] <- log(xlims[['EHMacute']]+10)
}
pmc0 <- addEHM(simParmSamp(4*10^4)) ## sample prior distribution
pmc0 <- within(pmc0, {for(ll in logParms) assign(paste0('log',ll), log(get(ll))); rm(ll)}) ## add log variables to priors
for(bb in 1:finalbatch) { ## load all batches as pmc1, ..., pmcN
load(file=file.path(out.dir, paste0('IntermedDistr',bb,'.Rdata'))) ## Load last distribution (already filtered)
assign(paste0('pmc',bb), pmatChosen) ## pmc = parameter matrix chosen from ABC, number indicates ABC batch (intermediate distribution)
}
adjustBW <- c(rep(2,5))
for(bb in 0:finalbatch) for(ii in 1:npars) {
pp <- parsDo[ii]
tempPmc <- get(paste0('pmc',bb))
if(logDo[ii]) {
temp <- with(tempPmc, density(log(get(pp)), from = rgs[[pp]][1], to = rgs[[pp]][2], adjust=adjustBW[ii]))
}else{
temp <- with(tempPmc, density(get(pp), from = rgs[[pp]][1], to = rgs[[pp]][2], adjust=adjustBW[ii]))
}
if(transfEHM & pp=='EHMacute') temp <- with(tempPmc, density(log(get(pp)+10), from = rgs[[pp]][1], to = rgs[[pp]][2], adjust=adjustBW[ii]))
assign(paste0('Dens',pp,bb), temp)
if(bb==0) assign(paste0('Max',pp), max(temp$y)) else assign(paste0('Max',pp), max(temp$y, get(paste0('Max',pp))))
}
## Show approach from prior to posterior through intermediate distributions for EHMacute, het.gen.sd,
toDo <- 1:npars
cols <- c('black',colorRampPalette(c('gray','yellow','red'))(finalbatch))
cols <- c('black',rainbow(finalbatch))
#toDo <- (1:npars)[-3]
for(ff in 1:2) {
if(ff==1) pdf(file.path(fig.dir, 'posterior approach EHM.pdf'), w = 6.83, h=5) else{
png(file.path(fig.dir, 'posterior approach EHM.png'), w = 6.83, h=5, units='in',res=200)}
par(mfrow=c(2,3), lwd=2, bty='n', mar = c(5,4,2,.5),'ps'=12)
for(pp in toDo) for(bb in 0:finalbatch) {
tempDens <- get(paste0('Dens',parsDo[pp],bb))
if(bb==0) {
plot(tempDens, xlab=nmsDo[pp], ylab = 'density', xlim = rgs[[pp]], type = ifelse(flatPriors[pp],'n','l'),
col=cols[bb+1], ylim = c(0,get(paste0('Max',parsDo[pp]))), las = 1, main = paste0('(',LETTERS[pp],')'), xaxt='n')
if(flatPriors[pp]) segments(rgs[[pp]][1], flatDens[[pp]],rgs[[pp]][2], flatDens[[pp]], col = cols[bb+1])
if(logDo[pp]) axis(1, at = logxticks[[pp]], label = xticks[[pp]]) else{
if(transfEHM & parsDo[pp]=='EHMacute') axis(1, at = ehmticks, label=xticks[[pp]]) else axis(1, at = xticks[[pp]])
}
}else{
lines(tempDens, col = cols[bb+1])
}
}
par(mar=rep(2,4))
plot(0,0, type = 'n', bty ='n', axes=F)
legend('topright', distrNms, col = cols, lwd = 2, bty = 'n', cex=.8)
}
graphics.off()
## Without EHMacute
toDo <- (1:npars)[-3]
for(ff in 1:2) {
if(ff==1) pdf(file.path(fig.dir, 'posterior approach.pdf'), w = 6.83, h=5) else{
png(file.path(fig.dir, 'posterior approach.png'), w = 6.83, h=5, units='in',res=200)}
par(mfrow=c(2,2), lwd=2, bty='n', mar = c(5,4,2,.5),'ps'=12)
for(ipp in 1:length(toDo)) for(bb in 0:finalbatch) {
pp <- toDo[ipp]
tempDens <- get(paste0('Dens',parsDo[pp],bb))
if(bb==0) {
plot(tempDens, xlab=nmsDo[pp], ylab = 'density', xlim = rgs[[pp]], type = ifelse(flatPriors[pp],'n','l'),
col=cols[bb+1], ylim = c(0,get(paste0('Max',parsDo[pp]))), las = 1, main = paste0('(',LETTERS[ipp],')'), xaxt='n')
if(flatPriors[pp]) segments(rgs[[pp]][1], flatDens[[pp]],rgs[[pp]][2], flatDens[[pp]], col = cols[bb+1])
if(logDo[pp]) axis(1, at = logxticks[[pp]], label = xticks[[pp]]) else{
if(transfEHM & parsDo[pp]=='EHMacute') axis(1, at = ehmticks, label=xticks[[pp]]) else axis(1, at = xticks[[pp]])
}
}else{
lines(tempDens, col = cols[bb+1])
}
}
if(ipp==4) legend('topleft', distrNms, col = cols, lwd = 2, bty = 'n', cex = .8)
}
graphics.off()
## CI convergence
for(bb in 0:finalbatch) {
tempArr <- quants(get(paste0('pmc',bb))[,c('EHMacute','acute.sc','dur.ac','het.gen.sd','bp')])
if(bb==0) ciArr <- tempArr else ciArr <- abind(ciArr,tempArr, along=3)
}
ciArr[,'EHMacute',] ## still shrinking in batch 3->4
ciArr ## still shrinking in batch 3->4
for(bb in 0:finalbatch) write.csv(ciArr[,,bb], file.path(fig.dir, paste0('ciArr',bb,'.csv')))
ciArr[,'bp',6]*12
## Get median transmission rates (i.e. mean on log scale) for paper
hazpost <- get(paste0('pmc',finalbatch))
hazpost <- within(hazpost, {
medianbp <- bp / exp(het.gen.sd^2 / 2) ## mean logbp is median bp
annualMedian_bp <- medianbp*12
annualMedian_bpAcute <- annualMedian_bp*acute.sc
annualMean_bp <- bp*12
bp025 <- exp(qnorm(.025,0,2))*annualMedian_bp
bp975 <- exp(qnorm(.975,0,2))*annualMedian_bp
})
quants(hazpost[,c('bp','het.gen.sd','logbp','medianbp','annualMedian_bp','annualMean_bp','annualMedian_bpAcute')])
varsGiven <- c(paste0('log',ghazs), paste0('log',mfs), 'logacute.sc','logdur.ac','het.gen.sd','medianbp','annualMedian_bp','annualMean_bp')
posteriorABCSMC <- hazpost[,varsGiven]
toReplace <- c('logacute.sc','logdur.ac','het.gen.sd','medianbp','annualMedian_bp','annualMean_bp')
colnames(posteriorABCSMC)[colnames(posteriorABCSMC) %in% toReplace] <- c('logRHacute','logDacute','sigma_hazard','median_lambda','annual_median_lambda', 'annual_mean_lambda')
head(posteriorABCSMC)
save(posteriorABCSMC, file=file.path(out.dir, 'Data File S1.Rdata'))
png(file.path(fig.dir, 'lambda vs sigma.png'), w = 4.5, h=5.5, units='in', res = 200)
smoothScatter(hazpost$annualMedian_bp, hazpost$het.gen.sd)
dev.off()
finalbatch <- 5
## Table of proportion of couples by interval from final distribution
load(file=file.path(out.dir, paste0('IntermedDistr',finalbatch,'.Rdata'))) ## Load last distribution (already filtered)
propFromWtab <- function(x) x[[1]][,3,] / apply(x[[1]][,3:4,],c(1,3),sum)
WpropsRl <- signif(with(wtab.rl, cbind(with(inct, i/n), with(prevt, i/n))),2)
Wprops <- lapply(gtabs, propFromWtab)
Wprops <- array(unlist(Wprops), dim=c(4,2,length(Wprops)))
WpropsMed <- signif(apply(Wprops, 1:2, function(x) quantile(x, .5, na.rm=T)),2)
WpropsLCI <- signif(apply(Wprops, 1:2, function(x) quantile(x, .025, na.rm=T)),2)
WpropsUCI <- signif(apply(Wprops, 1:2, function(x) quantile(x, .975, na.rm=T)),2)
cistrg <- function(med,lci,uci) paste0(med, ' (',lci,', ',uci,')')
WtabFit <- data.frame(WawInc=WpropsRl[,1], PostInc=cistrg(WpropsMed[,1],WpropsLCI[,1],WpropsUCI[,1]),
WawInc=WpropsRl[,2], PostInc=cistrg(WpropsMed[,2],WpropsLCI[,2],WpropsUCI[,2]))
write.csv(WtabFit, file.path(fig.dir, 'Fit to proportions in Wawer table.csv'))
WtabFit
for(ff in 1:2) {
if(ff==1) pdf(file.path(fig.dir, 'GoF.pdf'), w = 3.5, h=5.5) else{
png(file.path(fig.dir, 'GoF.png'), w = 3.5, h=5.5,units='in',res=200) }
par(mfcol=c(4,2), mar = c(2,2,2,.5), oma = c(3,3,0,0), 'ps'=10)
for(jj in 1:2) for(ii in 1:4) {
main <- ifelse(ii==1, c('incident couples','prevalent couples')[jj], '')
hist(Wprops[ii,jj,], breaks = ifelse(jj==1&ii==2,5,10), xlab = '', col = 'black', main = main, yaxt='n', xlim = c(0,1))
abline(v=WpropsRl[ii,jj], col = 'red')
if(jj==1) mtext(paste0('interval',ii) ,2, 3)
}
mtext('proportion seroconverting',1, 1, outer=T)
mtext('frequency',2, -1, outer=T)
graphics.off()
}
## Table of RHacute & dacute values for future modeling work
## Look at CIs nonlog
apply(pmatChosen[,c(parnms,'EHMacute')], 2, function(x) quantile(x,c(.025,.5,.975)))
with(pmatChosen, cor.test(EHMacute, het.gen.sd))
## How many simulations were run?
numSims <- rep(NA,finalbatch)
for(bb in 1:finalbatch) {
load(file=file.path(out.dir, paste0('pmatLs',bb,'.Rdata'))) ## Load last distribution (already filtered)
numSims[bb] <- sum(unlist(lapply(pmatLs, function(x) nrow(x$pmat))))
}
numSims
sum(numSims)*1.2/60/24