-
Notifications
You must be signed in to change notification settings - Fork 0
/
recovery_setup.R
386 lines (290 loc) · 13.8 KB
/
recovery_setup.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
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
# Clear workspace
rm(list = ls())
# Set working directory to top-level folder containing DMC
getwd()
# Load libraries and DMC functions
source("dmc/dmc.R")
load_model("LBA", "lbaN_B.R")
# source("dmc/dmc_ATC.R")
# source("dmc/dmc_PMDC.R")
# Load samples object
print(load("samples/sTPPM_full_sdvS.RData"))
samples <- samples1
colnames(samples[[1]]$theta)
# Set up parameters and priors --------------------------------------------
# Mean (here put your actual priors that you used to fit the model)
p.vector <- c(t0 = 0.3, A = 1.5,
B.3s10C = 2, B.3s30C = 2, B.6s10C = 2, B.6s30C = 2,
B.3s10N = 2, B.3s30N = 2, B.6s10N = 2, B.6s30N = 2,
B.3s10P = 2, B.3s30P = 2, B.6s10P = 2, B.6s30P = 2,
mean_v.cc3s10C = 1, mean_v.nn3s10C = 0, mean_v.pc3s10C = 0, mean_v.pn3s10C = 0,
mean_v.cc6s10C = 1, mean_v.nn6s10C = 0, mean_v.pc6s10C = 0, mean_v.pn6s10C = 0,
mean_v.cc3s30C = 1, mean_v.nn3s30C = 0, mean_v.pc3s30C = 0, mean_v.pn3s30C = 0,
mean_v.cc6s30C = 1, mean_v.nn6s30C = 0, mean_v.pc6s30C = 0, mean_v.pn6s30C = 0,
mean_v.cc3s10N = 0, mean_v.nn3s10N = 1, mean_v.pc3s10N = 0, mean_v.pn3s10N = 0,
mean_v.cc6s10N = 0, mean_v.nn6s10N = 1, mean_v.pc6s10N = 0, mean_v.pn6s10N = 0,
mean_v.cc3s30N = 0, mean_v.nn3s30N = 1, mean_v.pc3s30N = 0, mean_v.pn3s30N = 0,
mean_v.cc6s30N = 0, mean_v.nn6s30N = 1, mean_v.pc6s30N = 0, mean_v.pn6s30N = 0,
mean_v.pp3s10P = 1, mean_v.pp6s10P = 1, mean_v.pp3s30P = 1, mean_v.pp6s30P = 1,
mean_v.PMFA = 0,
sd_v.ccSDV = 0.5, sd_v.nnSDV = 0.5)
# Set priors
p.prior <- prior.p.dmc(
dists = c("beta", rep("tnorm", length(p.vector)-1)),
p1 = c(p.vector),
p2 = c(0.2, 0.1, rep(1, 12), rep(1, 37), rep(0.5, 2)),
lower = c(0.1, 0, rep(0, 12), rep(NA, 37), rep(0, 2)),
upper = c(1, 10, rep(Inf, 12), rep(Inf, 37), rep(Inf, 2))
)
length(p.prior)
length(p.prior) * 3
# Get the average amount of data per participant
Ns <- lapply(samples, get.ns.dmc)
Ns
# Get the average total number of trials per participant observed in the experiment
# (this accounts for reduction in power from removed RT outlier trials etc.)
ns <- ceiling(Reduce("+", Ns)/length(Ns))
ns
# Get model from samples object
model <- attr(samples[[1]]$data, "model")
attr(model, "match.map")
# Get representative parameters from the top model
# Get posterior mean (or median) parameters of the subject-average distribution
msds <- get.msds(samples)
msds <- data.frame(msds)
msds
sim_pop_mu <- msds$M; names(sim_pop_mu) <- rownames(msds)
sim_pop_mu
# Simulate synthetic data -------------------------------------------------
# Loop through and make 100 synthetic participants with the representative data frame size
for (i in 1:100) {
data <- simulate.dmc(sim_pop_mu, model, n = ns)
data <- cbind(i, data)
if (i == 1) okdats <- data else okdats <- rbind(okdats, data)
}
# Label subjects column 's'
head(okdats)
names(okdats)[1] <- "s"
okdats$s <-factor(okdats$s)
# Make data model
dm <- data.model.dmc(okdats, model)
# Sampling ----------------------------------------------------------------
# Settings
n.chains <- length(p.prior) * 3
# Generate start points
recov_samples <- h.samples.dmc(nmc = 100,
p.prior = p.prior,
data = dm,
thin = 10,
n.chains = n.chains)
# Save samples and run on servers with "batch" file (batch_<model_name>.R)
# with the command: nohup R CMD BATCH batch_<model_name>.R &
save(recov_samples, file = "samples/recov_sTPPM_full_sdvS.RData")
# Run fitting on grid with batch script
# -------------------------------------------------------------------------
# Model mimicry recovery study --------------------------------------------
#
# 1. Generate data from a model with certain effects
# deleted (by changing the p.vector you simulate from)
# 2. Fit the top model to that and check whether the
# deleted effects spontaneously emerge from nothing.
# If so, we should be suspicious of them.
# Setup environment -------------------------------------------------------
# Clear workspace
rm(list = ls())
# Set working directory
getwd()
# Load libraries and DMC functions
source("dmc/dmc.R")
load_model("LBA", "lbaN_B.R")
# Load top samples object
print(load("samples/sTPPM_full_sdvS.RData"))
top_samples <- samples1
colnames(top_samples[[1]]$theta)
# Set up parameters and priors --------------------------------------------
# Mean (here put your actual priors that you used to fit the model)
p.vector <- c(t0 = 0.3, A = 1.5,
B.3s10C = 2, B.3s30C = 2, B.6s10C = 2, B.6s30C = 2,
B.3s10N = 2, B.3s30N = 2, B.6s10N = 2, B.6s30N = 2,
B.3s10P = 2, B.3s30P = 2, B.6s10P = 2, B.6s30P = 2,
mean_v.cc3s10C = 1, mean_v.nn3s10C = 0, mean_v.pc3s10C = 0, mean_v.pn3s10C = 0,
mean_v.cc6s10C = 1, mean_v.nn6s10C = 0, mean_v.pc6s10C = 0, mean_v.pn6s10C = 0,
mean_v.cc3s30C = 1, mean_v.nn3s30C = 0, mean_v.pc3s30C = 0, mean_v.pn3s30C = 0,
mean_v.cc6s30C = 1, mean_v.nn6s30C = 0, mean_v.pc6s30C = 0, mean_v.pn6s30C = 0,
mean_v.cc3s10N = 0, mean_v.nn3s10N = 1, mean_v.pc3s10N = 0, mean_v.pn3s10N = 0,
mean_v.cc6s10N = 0, mean_v.nn6s10N = 1, mean_v.pc6s10N = 0, mean_v.pn6s10N = 0,
mean_v.cc3s30N = 0, mean_v.nn3s30N = 1, mean_v.pc3s30N = 0, mean_v.pn3s30N = 0,
mean_v.cc6s30N = 0, mean_v.nn6s30N = 1, mean_v.pc6s30N = 0, mean_v.pn6s30N = 0,
mean_v.pp3s10P = 1, mean_v.pp6s10P = 1, mean_v.pp3s30P = 1, mean_v.pp6s30P = 1,
mean_v.PMFA = 0,
sd_v.ccSDV = 0.5, sd_v.nnSDV = 0.5)
# Set priors
p.prior <- prior.p.dmc(
dists = c("beta", rep("tnorm", length(p.vector)-1)),
p1 = c(p.vector),
p2 = c(0.2, 0.1, rep(1, 12), rep(1, 37), rep(0.5, 2)),
lower = c(0.1, 0, rep(0, 12), rep(NA, 37), rep(0, 2)),
upper = c(1, 10, rep(Inf, 12), rep(Inf, 37), rep(Inf, 2))
)
length(p.prior)
length(p.prior) * 3
# Get the average amount of data per participant
Ns <- lapply(top_samples, get.ns.dmc)
Ns
# Get the average total number of trials per participant observed in the experiment
# (this accounts for reduction in power from removed RT outlier trials etc.)
ns <- ceiling(Reduce("+", Ns)/length(Ns))
ns
# Get model from your samples object
model <- attr(top_samples[[1]]$data, "model")
attr(model, "match.map")
# Get representative parameters from the top model
# Get posterior mean (or median) parameters of the subject-average distribution
top_msds <- get.msds(top_samples)
top_msds <- data.frame(top_msds)
top_msds
sim_pop_mu <- top_msds$M; names(sim_pop_mu) <- rownames(top_msds)
sim_pop_mu
# -------------------------------------------------------------------------
# No threshold effects recovery model -------------------------------------
#
# Here we use the model fitted without threshold effects as the source
# of the parameters to get parameter values that actually could have fit
# our study without threshold effects
#
# Load samples object from reduced model
print(load("samples/sTPPM_v_sdvS.RData"))
replacement_samples <- samples1
colnames(replacement_samples[[1]]$theta)
# Get replacement parameters
replacement_msds <- get.msds(replacement_samples) # Reduced model
replacement_pop_mu <- data.frame(replacement_msds)$M
names(replacement_pop_mu) <- rownames(replacement_msds)
replacement_pop_mu
sim_pop_mu
length(replacement_pop_mu)
length(sim_pop_mu)
# Replace full model parms with reduced model parms
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^A")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^A")] # A
# Conflict thresholds
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^B.*C")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^B.*C")]
# Non-conflict thresholds
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^B.*N")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^B.*N")]
# PM thresholds
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^B.*P")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^B.*P")]
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^t0")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^t0")] # t0
# Rates
# names(sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v")]) ==
# names(replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v")])
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v")]
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^sd_v")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^sd_v")] # sdv
save(sim_pop_mu, file = "deriv/recov_p_vector_v_sdvS.RData")
# Simulate synthetic data -------------------------------------------------
# Loop through and make 100 synthetic participants with the representative data frame size
for (i in 1:100) {
data <- simulate.dmc(sim_pop_mu, model, n = ns)
data <- cbind(i, data)
if (i == 1) okdats <- data else okdats <- rbind(okdats, data)
}
# Label subjects column 's'
head(okdats)
names(okdats)[1] <- "s"
okdats$s <-factor(okdats$s)
# Make data model
dm <- data.model.dmc(okdats, model)
# Sampling ----------------------------------------------------------------
# Settings
n.chains <- length(p.prior) * 3
# Set up a samples (start points) object to fit to the synthetic data
recov_samples <- h.samples.dmc(nmc = 100,
p.prior = p.prior,
data = dm,
thin = 10,
n.chains = n.chains
)
save(recov_samples, file = "samples/recov_sTPPM_v_sdvS.RData")
# Run fitting on grid with batch script
# -------------------------------------------------------------------------
# No drift rate effects recovery model ------------------------------------
#
# Use the model fitted without drift rate effects as the source of the
# parameters to get parameter values that actually could have fit our
# study without drift rate effects
#
# Load samples object from reduced model
print(load("samples/sTPPM_B_sdvS.RData"))
replacement_samples <- samples1
colnames(replacement_samples[[1]]$theta)
# Get replacement parameters
replacement_msds <- get.msds(replacement_samples) # Reduced model
replacement_pop_mu <- data.frame(replacement_msds)$M
names(replacement_pop_mu) <- rownames(replacement_msds)
replacement_pop_mu
sim_pop_mu
length(replacement_pop_mu)
length(sim_pop_mu)
# Replace full model parameters with reduced model parameters
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^A")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^A")] # A
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^B")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^B")] # B
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^t0")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^t0")] # t0
# Conflict rates
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v.cc.*C")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v.cc.*C")]
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v.nn.*C")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v.nn.*C")]
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v.pc.*C")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v.pc.*C")]
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v.pn.*C")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v.pn.*C")]
# Non-conflict rates
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v.cc.*N")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v.cc.*N")]
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v.nn.*N")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v.nn.*N")]
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v.pc.*N")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v.pc.*N")]
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v.pn.*N")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v.pn.*N")]
# PM rates
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v.pp.*P")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v.pp.*P")]
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^mean_v.PMFA")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^mean_v.PMFA")]
sim_pop_mu[grep(names(sim_pop_mu), pattern = "^sd_v")] <-
replacement_pop_mu[grep(names(replacement_pop_mu), pattern = "^sd_v")] # sdv
save(sim_pop_mu, file = "deriv/recov_p_vector_B_sdvS.RData")
# Simulate synthetic data -------------------------------------------------
# Loop through and make 100 synthetic participants with the representative data frame size
for (i in 1:100) {
data <- simulate.dmc(sim_pop_mu, model, n = ns)
data <- cbind(i, data)
if (i == 1) okdats <- data else okdats <- rbind(okdats, data)
}
# Label subjects column 's'
head(okdats)
names(okdats)[1] <- "s"
okdats$s <-factor(okdats$s)
# Make data model
dm <- data.model.dmc(okdats, model)
# Sampling ----------------------------------------------------------------
# Settings
n.chains <- length(p.prior) * 3
# Set up a samples (start points) object to fit to the synthetic data
recov_samples <- h.samples.dmc(nmc = 100,
p.prior = p.prior,
data = dm,
thin = 10,
n.chains = n.chains
)
save(recov_samples, file = "samples/recov_sTPPM_B_sdvS.RData")
# Run fitting on grid with batch script