-
Notifications
You must be signed in to change notification settings - Fork 1
/
B4_model_predictions_power_hierarchical_2alpha.R
345 lines (257 loc) · 9.82 KB
/
B4_model_predictions_power_hierarchical_2alpha.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
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
# Summary stats of model predictions from run of Bayes water quality model
#
# CJ Brown 17 May 2016
# This version uses a power function and source contributions are sampled from a heirarchical model.
rm(list = ls())
library(splines)
library(RColorBrewer)
library(ggplot2)
library(tidyr)
library(corrplot)
library(rjags)
setwd('/Users/s2973410/Databases/Fiji reefs analysis/wq_models/large_region_data_v3')
source('~/Code/Useful R Code/figlabel.R')
# runname <- 'large_region_hp_21Mar16_unscaled'
# runname <- 'large_region_2Mar16_exp_v3'
# runname <- 'large_region_hp_1Apr16_unscaled'
# runname <- 'large_region_hp_3Apr16_unscaled'
# runname <- 'large_region_hp_28Apr16'
# runname <- 'large_region_hp_17May16_2alpha'
runname <- 'large_region_hp_24May16_2alpha'
# dathighres <- read.csv('WQ_dat_highres_grouped.csv', header = T)
# save(list = 'dathighres', file = 'WQ_dat_highres_grouped.RData')
load('WQ_dat_highres_grouped.RData') #use load to save some time
icellhr <- read.csv('sampled_cell_ids.csv')$x
#Validation data
ikeep <- which(!dathighres$cells %in% icellhr)
datv <- dathighres[ikeep,]
#Check validation data
# datv2 <-dathighres[which(dathighres$cells %in% icellhr),]
# library(sp)
# datsp <- datv[,c(30:32)]
# datsp2 <- datv2[,c(30:32)]
# coordinates(datsp) <- ~x+y
# coordinates(datsp2) <- ~x+y
# plot(datsp)
# plot(datsp2, add = T, col ='red')
rm(dathighres)
#
# Some key catchments
#
icatch <- c(8, 15, 7, 12, 23, 10, 26, 27)
# catchnams <- c('Savusavu Bay catchments','Dreketi river',
# 'Bua Bay catchments', 'Dawara river' ,'Labasa', 'Wainunu Bay catchments')
catchnams <- c('','Dreketi river',
'Bua Bay catchments', '' ,'Labasa', '', 'Nasauu river','Rukuruku Bay')
#
# Load last run
#
load(paste(runname,'RData', sep ='.'))
#
# Data summaries
#
smc <- summary(mcout)
smc
#Which rows of summary are which params?
dimnam <- attr(smc$stat, 'dimnames')[[1]]
ibeta <- grep('lnbeta', dimnam)
ialpha <- grep('alpha', dimnam)
ilnb <- grep('lnb', dimnam)[1]
a_est <- -smc$statistics[ialpha,1]
b_est <- exp(smc$statistics[ilnb,1])
betaest <- exp(smc$statistics[ibeta,1])
betaCI <- exp(smc$quantiles[ibeta,c(1,5)])
sdest <- sqrt(1/smc$statistics[grep('tau', dimnam),1][1])
#
# Plot traces for alpha and ascale
#
dev.new()
par(mfrow =c(2,3))
plot(as.numeric(mcout[[1]][,ialpha[1]]),type = 'l', xlab = '', ylab ='', main = 'alpha')
plot(as.numeric(mcout[[1]][,ialpha[2]]),type = 'l', xlab = '', ylab ='', main = 'alpha')
plot(as.numeric(mcout[[1]][,ilnb]),type = 'l', xlab = '', ylab ='', main = 'ln(w)')
plot(as.numeric(mcout[[1]][,ialpha[1]]), as.numeric(mcout[[1]][, ilnb]),type = 'p', xlab = '', ylab ='', main = 'correlation')
plot(as.numeric(mcout[[1]][,ialpha[2]]), as.numeric(mcout[[1]][, ilnb]),type = 'p', xlab = '', ylab ='', main = 'correlation')
#
# Plot betas vs yields
#
pchcols <- grey(0.2, 0.4)
figlab <- LETTERS[1:4]
xoff <- 0.05
yoff <- 0.08
letcex <- 0.6
letcex1 <- 1.5
#Offsets for labels
xofflab <- rep(-0.03, length(catchnams))
xofflab[8] <- -0.1
yofflab <- c(-0.04, 0, 0.1, 0.02, 0, 0, 0, -0.15)
dev.new(width = 9, height = 4)
par(mfrow = c(1,2), mar = c(5,5,2,2))
plot(bmean, betaest/b_est, col = pchcols, ylim = c(0,4), xlim = c(0,1.5), pch = 16, xlab = 'Relative influence on turbidity', ylab = 'Relative influence on turbidity (posterior)')
arrows(bmean, betaCI[,1]/b_est, bmean, betaCI[,2]/b_est, len =0, col = pchcols)
abline(0, 1, col = 'grey80')
text(bmean[icatch]+xofflab, (betaest/b_est)[icatch]+yofflab, catchnams, cex = letcex, pos = 4)
# text(bmean, betaest/b_est, 1:28)
figlabel(label = figlab[1], font =2, xoffset=xoff, yoffset = yoff, cex =letcex1)
plot(log(bmean), log(betaest/b_est), ylim = c(-14, 2), xlim = c(-10, 0), pch = 16, col = pchcols, xlab = 'Relative influence on turbidity (log)', ylab = 'Relative influence on turbidity (posterior, log)')
arrows(log(bmean), log(betaCI[,1]/b_est), log(bmean), log(betaCI[,2]/b_est), len =0, col = pchcols)
abline(0,1, col = 'grey80')
figlabel(label = figlab[2], font =2, xoffset=xoff, yoffset = yoff, cex =letcex1)
#Plot for presentation
dev.new()
par(mar =c(5,6,4,2))
plot(log10(bmean), log10(betaest/b_est), ylim = c(-6, 2), xlim = c(-3, 0),
pch = 16, col = pchcols, cex = 1.5, xlab = "", ylab = "", cex.axis = 2, las = 1, xaxt = 'n', yaxt = 'n')
#arrows(log10(bmean), log10(betaCI[,1]/b_est), log10(bmean), log10(betaCI[,2]/b_est), len =0, col = pchcols,
#lwd = 2)
abline(0,1, col = 'blue', lwd = 2)
axis(1, cex = 1.5, at = -3:0, labels = 10^(-3:0), cex.axis = 2)
axis(2, cex = 1.5, at = seq(-6, 2, by = 2), labels = 10^seq(-6, 2, by = 2), cex.axis = 2, las = 1)
dev.new()
par(mar =c(5,6,4,2))
plot(b1$mnyld_tons, as.numeric(betaest/b_est), ylim = c(0.000001, 150),
xlim = c(100, 65000), pch = 16, col = pchcols,
xlab = '', ylab = '', log = 'xy', las = 1, cex.axis = 2, cex = 1.5)
arrows(b1$mnyld_tons, betaCI[,1]/b_est, b1$mnyld_tons, betaCI[,2]/b_est, len =0, col = pchcols, lwd = 2)
# text(bmean, betaest, 1:28)
#
# Plot correlation between parameters
#
cor(mcout[[1]])
corrplot(cor(mcout[[1]]), diag = F, type = 'lower')
mcall <- rbind(mcout[[1]], mcout[[2]], mcout[[3]])
corrplot(cor(mcall), diag = F, type = 'lower')
#
# Convergence
#
gelman.diag(mcout, autoburnin = F)
#
# Predictions
#
bmat <- matrix(rep(betaest, length(y)), ncol = nriv, byrow = T)
amat <- matrix(rep(a_est[icoast],length(y)), ncol = nriv, byrow = T)
ymat <- bmat *(xdat^ amat)
ypred <- rowSums(ymat)*exp(sdest/2)
yresid <- yscale - ypred
cor(ypred, yscale)
cor(log(ypred), lny)
plot(log(ypred), lny, xlab = 'predictions', ylab = 'observed'); abline(0,1)
plot(ypred, yscale, xlab = 'predictions', ylab = 'observed', xlim = c(0,1)); abline(0,1)
plot(ypred, yresid, xlab = 'predictions', ylab = 'residuals'); abline(0,0)
#
# Predictive deviance
#
lnypred <- log(ypred)
Gm <- sum((lnypred - lny)^2)
Pm <- mean(lnypred ^2) - (mean(lnypred)^2)
var(lny - lnypred)
var(lnypred)
Gm + Pm
#
# Calculate DIC
#
predictmodel <- function(betavalf, yf, nrivf, xmapf, a_estf, icoastf){
# Create predictions df
bmapf <- matrix(rep(betavalf, length(yf)), ncol = nrivf, byrow = T)
amapf <- matrix(rep(a_estf[icoastf], length(yf)), ncol = nrivf, byrow = T)
log(rowSums(bmapf * (xmapf ^ amapf)))
}
predictmodel_mean <- function(betavalf, yf, nrivf, xmapf, a_estf, icoastf, sdest){
# Create predictions df
bmapf <- matrix(rep(betavalf, length(yf)), ncol = nrivf, byrow = T)
amapf <- matrix(rep(a_estf[icoastf], length(yf)), ncol = nrivf, byrow = T)
log(rowSums(bmapf * (xmapf ^ amapf))*exp(sdest/2))
}
nchain <- length(mcout)
nstep <- nrow(mcout[[1]])
dimnam <- attr(mcout[[1]], 'dimnames')[[2]]
ibeta <- grep('lnbeta',dimnam)
ialpha <- grep('alpha',dimnam)
itaubeta <- grep('taubeta',dimnam)
itau <- grep('tau',dimnam)
itau <- itau[itau!=itaubeta]
# Posterior deviance for each step
loglik <- matrix(NA, nrow = nstep, ncol = nchain)
for (ichain in 1:nchain){
xchain <- as.matrix(mcout[[ichain]])
for (i in 1:nstep){
lnypred <- predictmodel(exp(xchain[i,ibeta]), y, nriv, xdat, -xchain[i, ialpha], icoast)
loglik[i, ichain] <- sum(-2*dnorm(lny, mean = lnypred, sd = sqrt(1/xchain[i,itau]), log = T))
}
}
# Posterior averaged deviance
lnypred <- predictmodel_mean(betaest, y, nriv, xdat, a_est, icoast, sdest)
loglik_bar <- sum(2*dnorm(lny, mean = lnypred, sd = sdest, log = T))
# Effective number of params
pD <- mean(loglik) + loglik_bar
# DIC
mean(loglik) + pD
#
# Predict to cells not used in fitting
#
yobshr <- as.numeric(datv$wq)
preddat <- as.numeric(unlist(datv[,icol]))
preddat <- matrix(preddat, ncol = nriv, byrow = F)*xscale
betaest2 <- betaest
# betaest2[1] <- 1
bmathr <- matrix(rep(betaest2, length(yobshr)), ncol = nriv, byrow = T)
amathr <- matrix(rep(a_est[icoast], length(yobshr)), ncol = nriv, byrow = T)
ymathr <- bmathr *(preddat^ amathr)
ypredhr <- rowSums(ymathr)*exp(sdest/2)
yntuhr <- ypredhr + (ymin-cadd)
yresidhr <- yobshr - yntuhr
plot(yntuhr, yresidhr); abline(0,0)
plot(log(yntuhr), log(yobshr)); abline(0,1)
cor(yobshr, yntuhr)
cor(log(yntuhr), log(yobshr))
# Save df of high res data
dfsav <- data.frame(predntu = yntuhr, obsntu = yobshr)
write.csv(dfsav, paste0(runname, 'datpred_highres.csv'))
#Root mean-squared error - error in NTUs
#This equals sd if there is no bias
sqrt(sum((yntuhr - yobshr)^2)/length(yntuhr))
rm(preddat)
rm(bmathr)
rm(ymathr)
#
# Create plot of predictions versus distances
#
nx <- 100
xdist <- seq(1, 10, length.out = nx)
y1 <- betaest[10] * (xdist ^ a_est[icoast[10]])
y2 <- betaest[19] * (xdist ^ a_est[icoast[10]])
par(las = 1, mar = c(6,6,2,2))
plot(xdist, y1, ylim = c(0, 4), type = 'l', col = 'tomato3', lwd = 5, yaxt = 'n', xlab = 'Distance from river mouth (km)', cex.axis =2.53, tck = 0.02, cex.lab = 2.8, ylab = 'Contribution to turbidity', bty = 'l')
lines(xdist, y2, lwd = 5, col = 'steelblue')
lines(xdist, y1+y2, lwd = 5, col = 'black')
#
# Save summary data
#
save(list = ls(), file = paste(runname, '_stats','.RData', sep =''))
#
#Plot posterior for beta
#
library(ggplot2)
library(tidyr)
nrows <- length((mcout[[1]][,1]))
nchains <- length(mcout)
ybeta <- data.frame(matrix(NA, nrow = nrows * nchains, ncol = nriv-1))
for (i in 1:nriv){
for (j in 1:nchains){
ybeta[(j*nrows - nrows+1) : (j*nrows),i] <-
exp(as.numeric(mcout[[j]][,ibeta[i]]))
}
}
colnames(ybeta) <- paste('Source', 1:nriv, sep ='_')
ybetas <- gather(ybeta, betavar, Beta)
blines <- data.frame(betavar = unique(ybetas$betavar), bpost = betaest[1:nriv], bprior = bmean[1:nriv]*b_est)
ybetas$sourcenum <- as.numeric(unlist(lapply(strsplit(ybetas$betavar, '_'), function(x) x[2])))
ybetas <- ybetas[order(ybetas$sourcenum),]
ggplot(ybetas, aes(x=Beta)) +
geom_density(fill = 'grey', alpha = 0.2, color = 'grey') +
ylab("Density") +
facet_wrap( ~ betavar,nrow = 5, scales = 'free_y') +
theme_bw() +
geom_vline(data = blines, aes(xintercept = bpost), color = 'grey20') +
geom_vline(data = blines, aes(xintercept = bprior), color = 'grey20', lty = 2) +
theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank())