set.seed(42)

library("lme4")
library("eyetrackingR")

data("word_recognition")
data <- make_eyetrackingr_data(word_recognition, 
                               participant_column = "ParticipantName",
                               trial_column = "Trial",
                               time_column = "TimeFromTrialOnset",
                               trackloss_column = "TrackLoss",
                               aoi_columns = c('Animate','Inanimate'),
                               treat_non_aoi_looks_as_missing = TRUE)

# subset to response window post word-onset
response_window <- subset_by_window(data,
                                    window_start_time = 15500,
                                    window_end_time = 21000,
                                    rezero = FALSE)

# remove trials with > 25% of trackloss
response_window_clean <- clean_by_trackloss(data = response_window,
                                            trial_prop_thresh = .25)

# pick threshold t based on alpha = .05 two tailed
num_sub = length(unique((response_window_clean$ParticipantName)))
threshold_t = qt(p = 1 - .05/2, 
                 df = num_sub-1)

# create Target condition column
response_window_clean$Target <- as.factor( ifelse(test = grepl('(Spoon|Bottle)', response_window_clean$Trial), 
                                                  yes = 'Inanimate', 
                                                  no  = 'Animate'))

# create time sequence data (within-subject)
response_time <- make_time_sequence_data(response_window_clean,
                                         time_bin_size = 100, 
                                         predictor_columns = c("Target"),
                                         aois = "Animate",
                                         summarize_by = "ParticipantName")

# look for initial clusters using lmer (within-subject)
df_timeclust <- make_time_cluster_data(response_time, 
                                       test= "lmer",
                                       predictor_column = "Target",
                                       treatment_level = "Inanimate",
                                       threshold = threshold_t,
                                       formula = Prop ~ Target + (1 | ParticipantName))

# run permutations (within-subject)
clust_analysis <- analyze_time_clusters(df_timeclust,
                                        formula = Prop ~ Target + (1 | ParticipantName),
                                        within_subj = T,
                                        samples = 100)

# create time sequence data (between-subject)
response_time_between <- make_time_sequence_data(response_window_clean,
                                                 time_bin_size = 100, 
                                                 predictor_columns = c("Sex", "Trial"),
                                                 aois = "Animate",
                                                 summarize_by = "ParticipantName" )

# look for initial clusters using lmer (between-subject)
df_timeclust_between <- make_time_cluster_data(response_time_between, 
                                               test= "lmer",
                                               predictor_column = "Sex",
                                               treatment_level = "M",
                                               threshold = threshold_t,
                                               formula = Prop ~ Sex + (1 | Trial))
  # using (1 | ParticipantName) doesn't work:
  # 'number of levels of each grouping factor must be < number of observations'

# run permutations (between-subject) with within_subj set to T and F
clust_analysis_between.wsT <- analyze_time_clusters(df_timeclust_between,
                                                    formula = Prop ~ Sex + (1 | Trial),
                                                    within_subj = T,
                                                    samples=100)
(summary(clust_analysis_between.wsT))
## Test Type:	 lmer 
## Predictor:	 Sex 
## Formula:	 Prop ~ Sex + (1 | Trial) 
## Null Distribution   ====== 
##   Mean:		 -4.4887 
## 2.5%:		 -4.4887 
## 97.5%:		 -4.4887 
## Summary of Clusters ======
##   Cluster Direction SumStatistic StartTime EndTime Probability
## 1       1  Negative    -4.488748     15500   15700           1
clust_analysis_between.wsF <- analyze_time_clusters(df_timeclust_between,
                                                    formula = Prop ~ Sex + (1 | Trial),
                                                    within_subj = F,
                                                    samples=100)
## Error in analyze_time_bins.time_sequence_data: 
## The term 'Sex' was not found in your model.
## This can happen if your predictor is treatment-coded, in which case you should specify the `treatment_level` argument.
## For example, if 'Target' is treatment coded, then it becomes 'TargetInanimate'in the model, and you should set `treatment_level = 'Inanimate'`.
## The terms in the model were: '(Intercept)', 'SexM'

# look for initial clusters using lmer (between-subject)
df_timeclust_between_lm <- make_time_cluster_data(response_time_between, 
                                                  test= "lm",
                                                  predictor_column = "Sex",
                                                  treatment_level = "M",
                                                  threshold = threshold_t,
                                                  formula = Prop ~ Sex)

# run permutations (between-subject) with within_subj set to T and F
clust_analysis_between_lm.wsT <- analyze_time_clusters(df_timeclust_between_lm,
                                                       formula = Prop ~ Sex,
                                                       within_subj = T,
                                                       samples=100)
(summary(clust_analysis_between.wsT))
clust_analysis_between_lm.wsF <- analyze_time_clusters(df_timeclust_between_lm,
                                                       formula = Prop ~ Sex,
                                                       within_subj = F,
                                                       samples=100)