cse_simulation/cse_simul.R at master · mibognar/cse_simulation · GitHub

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
simulate_data = function(condition_parameters_data, participant_number, trial_number){
  library(readr)
  library(dplyr)
  library(tidyr)
  library(purrr)
  library(EZ2)
  library(fdrtool)
  library(ez)
  library(rtdists)
  library(gamlss.dist)
  library(lme4)
  library(lmerTest)
  library(parallel)

  warnings()
### Create participants
  source_data = data.frame(participant_id = c(1:participant_number),
                         rt_intercept = rnorm(participant_number, 0, 0.1), # A 100ms SD random variability in overall RT
                         congruency_random_slope = rnorm(participant_number, 0, .02), # A 20ms SD noise in congruency effect
                         interaction_random_slope = rnorm(participant_number, 0, .01)) %>% # A 10ms SD noise in congruency sequence effect
  uncount(2) %>%
  mutate(is_congruent = rep(0:1, each=1, length.out=n())) %>%
  uncount(2) %>%
  mutate(prev_congruent = rep(0:1, each=1, length.out=n())) %>%
  left_join(condition_parameters_data, by=c("is_congruent","prev_congruent")) %>% # Read global condition means based on the input data
  mutate(congruency_random_slope = case_when(is_congruent == 1 ~ congruency_random_slope,
                                     T ~ -congruency_random_slope), #  noise depending on current congruency
         interaction_random_slope = case_when(
           is_congruent == 1 & prev_congruent == 1 ~ interaction_random_slope,
           is_congruent == 1 & prev_congruent == 0 ~ -interaction_random_slope,
           is_congruent == 0 & prev_congruent == 1 ~ -interaction_random_slope,
           is_congruent == 0 & prev_congruent == 0 ~ interaction_random_slope, #  noise depending on previous and current congruency condition
           TRUE ~ 0
         )) %>%
  mutate(mean_rt = glob_rtm+rt_intercept + congruency_random_slope + interaction_random_slope) # participant conditional RT modulated by systematic noise

### Create diffusion model parameters
diffusion_data <- source_data %>%
  mutate(Result = pmap(list(
    ifelse(glob_rtc==1, 1-1/(N*2),glob_rtc),
    glob_rtv,
    mean_rt,
    s=1),
    Data2EZ),
    v = map_dbl(Result, "v"),
    a = map_dbl(Result, "a"),
    Ter = map_dbl(Result, "Ter")) %>%
  dplyr::select(-Result)

### Generate trials based on model parameters
testdf = diffusion_data %>%
  mutate(Ter = ifelse(Ter<.1, .1, Ter)) %>%
  mutate(Result=pmap(
    list(trial_number,
         a=a,
         v=v,
         t0=Ter),
    rdiffusion)) %>%
  unnest(Result)

return(testdf)
}

test_simulation = function(condition_parameters_data, participant_number, trial_number, sd_filter, flag) {
  options(scipen = 999)
  options(dplyr.summarise.inform = FALSE)
  testdf = simulate_data(condition_parameters_data, participant_number, trial_number)

  ## Filtering data

  participant_summary = testdf %>%
    mutate(correct = ifelse(response == "upper", 1, 0)) %>%
    group_by(participant_id, is_congruent, prev_congruent) %>%
    summarize(N = n(),
              participant_mean_rt = mean(rt),
              participant_var_rt = var(rt),
              participant_sd_rt = sd(rt),
              participant_correct_percent = mean(correct)) %>%
    ungroup()

  diffusion_parameters = participant_summary %>%
    mutate(Result = pmap(list(
      ifelse(participant_correct_percent==1, 1-1/(N*2),participant_correct_percent),
      participant_var_rt,
      participant_mean_rt,
      s=1),
      Data2EZ),
      v = map_dbl(Result, "v"),
      a = map_dbl(Result, "a"),
      Ter = map_dbl(Result, "Ter")) %>%
    dplyr::select(-Result) %>%
    mutate(is_congruent = as.factor(is_congruent),
           prev_congruent = as.factor(prev_congruent),
           participant_id = as.factor(participant_id))

  testdf = testdf %>%
    left_join(participant_summary, by = c("participant_id", "is_congruent", "prev_congruent")) %>%
    mutate(rt_zscore = ((rt-participant_mean_rt)/participant_sd_rt))

  testfilter = testdf %>%
    filter(response == "upper",
           abs(rt_zscore)<sd_filter)

  #Fit model with glmer
  generalized_big_csemodel = glmer(rt ~ is_congruent*prev_congruent + (1+is_congruent*prev_congruent|participant_id), data = testfilter, family = inverse.gaussian(link = "log"),control=glmerControl(optimizer="bobyqa", optCtrl=list(maxfun=100000)))
  generalized_big_csemodel_summary = summary(generalized_big_csemodel)
  generalized_big_csemodel_p =generalized_big_csemodel_summary$coefficients[16]

  # Fit full linear model
  full_csemodel = lmer(rt ~ is_congruent*prev_congruent + (1+is_congruent*prev_congruent | participant_id), data = testfilter, control=lmerControl(optimizer="bobyqa", optCtrl=list(maxfun=100000)))
  full_csemodel_summary = summary(full_csemodel)
  full_csemodel_estimate = full_csemodel_summary$coefficients[4]*1000
  full_csemodel_p = full_csemodel_summary$coefficients[20]

  # Fit model with  intercept
  small_csemodel = lmer(rt ~ is_congruent*prev_congruent + (1|participant_id),data=testfilter, control=lmerControl(optimizer="bobyqa", optCtrl=list(maxfun=100000)))
  small_csemodel_summary = summary(small_csemodel)
  small_csemodel_p = small_csemodel_summary$coefficients[20]

  #Fit ANOVA RT

  csenova <- ezANOVA(data = diffusion_parameters,
                     dv = participant_mean_rt,
                     wid = participant_id,
                     within = .(is_congruent, prev_congruent)
  )
  cse_table = testfilter %>%
    group_by(prev_congruent, is_congruent) %>%
    summarize(mean_rt = mean(rt, na.rm=T)) %>%
    ungroup() %>%
    pivot_wider(names_from = c("prev_congruent","is_congruent"), names_sep = "_", values_from = mean_rt)

  cse = (cse_table$`1_0`-cse_table$`1_1`)-(cse_table$`0_0`-cse_table$`0_1`)>0
  anova_p = csenova$ANOVA$p[3]


  return(c(full_csemodel_estimate,
           ifelse(cse,generalized_big_csemodel_p<.05, F),
           ifelse(cse,full_csemodel_p<.05,F),
           ifelse(cse,small_csemodel_p<.05,F),
           ifelse(cse,anova_p<.05,F),
           flag))
  }

test_sequences = function(effect_table, runs, participants, trials){

  sd2list = vector("list", runs)
  sd3list = vector("list", runs)
  nonelist = vector("list", runs)
  for(j in 1:runs){
    sd2list[[j]] = test_simulation(effect_table, participants,trials,2.5, "2.5SD filter")
    print(j)
  }
  sd2df = as.data.frame(do.call(rbind, sd2list))

  for(j in 1:runs){
    sd3list[[j]] = test_simulation(effect_table, participants,trials,3, "3 SD filter")
    print(j)
  }
  sd3df = as.data.frame(do.call(rbind, sd3list))

  for(j in 1:runs){
    nonelist[[j]] = test_simulation(effect_table, participants,trials,1000, "No filter")
    print(j)
  }
  nonedf = as.data.frame(do.call(rbind, nonelist))

  alldf = rbind(sd2df,sd3df,nonedf)
  colnames(alldf) = c("estimate","gen_slope_p","slope_p", "intercept_p", "rt_anova_p","filtering")

  testsummary = alldf %>%
    group_by(filtering) %>%
    summarize(glmm_intercept_slope =mean(as.integer(as.logical(gen_slope_p)),na.rm=T),
              full_melr = mean(as.integer(as.logical(slope_p)), na.rm=T),
              intercept_melr = mean(as.integer(as.logical(intercept_p)), na.rm=T),
              anova_rt = mean(as.integer(as.logical(rt_anova_p)), na.rm=T)
              )

  return(testsummary)
}