circular_shift_bootstrap_functions.R

setwd('C:/cmatlab-Roman/DivPrograms/Individual/RomanDoronin/Scripts/Analysis')  
library(ggplot2)
library(plotly)
source("analyze_perches.R")
source("upload_s7s8.R")
source("ggplot-extensions.R")
path = "C:/cmatlab-Roman/DivPrograms/Individual/RomanDoronin/Data/s7s8/"
experiment = "s7s8"

#---- functions for the circular shifts plots  BEGIN  ----


MakeCircularShifts = function(bird_df){
  # Creates dataframe with randomly circularly shifted points within each perch window
  # For each perch it generates random amount of time shift and then we shift all elements within this perch on this amount
  #
  # Args: 
  #  bird_df: processed dataframe (note that you need columns t1 and t2 for each element (function Preprocessing does this) which says what are the timestamps of corresponding perch)
  # 
  # Returns:
  #  bird_df_shifted: identical by structure dataframe, but with modified timestamps
  perch_onsets = unique(bird_df$t1) # Creates a vector of onsets of the perches where there is at least one element from bird_df 
  indices = match(perch_onsets, bird_df$t1)
  perch_offsets = bird_df$t2[indices] 
  perch_durations = (perch_offsets - perch_onsets) * 3600 # Creates vector of durations of perches in order to generate random number from distribution, in sec - because we need to feed it to DFCircularShifts
  sample_shifts = sapply(perch_durations, FUN = function(x) runif(1, 0, x))
  map_df = data.frame(x = perch_onsets, y = sample_shifts)
  shift_times = map_df$y[match(bird_df$t1, map_df$x)]  # this is vector of time shifts for every element! (for each element within same perch window there is a same shift)
  bird_df_shifted = DFCircularShifts(bird_df, shift_times)
  return(bird_df_shifted)
}

DFCircularShifts = function(bird_df_mapped, shift_time) {
  # Makes circular shifts of the timestamps of the elements within their own communication window. 
  # Note: here as 
  #
  # Args:
  #  bird_df_mapped: processed dataframe. It is normal df, with added columns "window", "t1", "t2"
  #   "window": shows whether this element inside or outside perch window (need it in case if we have outsiders due to flatclust bug)
  #    "t1" - timestamp of the beginning of the communication window where this element was pronounced
  #    "t2" - timestamp of the end of the communication window where this element was pronounced
  #     this df can be generated by 
  #  shift_t: time of shift in seconds - vector! It says to what amount each element should shift! these times can be different for different elements!
  #
  # Returns:
  # input df, but with circularly shifted values
  
  #bird_df_mapped_filtered = filter(bird_df_mapped, window == "inside")  # if we will manage to fix the buf with outsiders, we won't 
                                                                        # need this line because all points will be inside window by definition
  time_h_shifted = mapply(function(t, t1, t2, shift_t) t1 + ((t + shift_t / 3600 - t1) %% (t2 - t1)),  # NOTE!!! Here I provide shift_t as a vector!!! 
                          bird_df_mapped$time_h, bird_df_mapped$t1, bird_df_mapped$t2, shift_time)  # You should feed good shift-t vector (how should we shift each element and 
  bird_df_mapped$time_h = time_h_shifted                                                                 # for elements within same perch window these should be the same!!)
  return(bird_df_mapped)
}






# Not needed really
# PlotCircularShifts = function(juvenile_df, adult_df, window_length=5, ...){
#   # Code for plotting peristimulus time histogram with circular shifts on the background
#   # Note: very buggy, better to rewrite (the main crap - is how to pass arguments to internal functions because different 
#   #       functions require different arguments)
#   #       also in seq(-5, 5, 0.5) there are magic numbers, need to rewrite with variables
#   p = CreateCumulativePlot(juvenile_df, adult_df, dt=0.15, ...)
#   for (i in seq(-5, 5, 0.5)){
#     shifted_adult = DFCircularShifts(adult_df, i)
#     temp_stack_df = DFForStackPlot(juvenile_df, shifted_adult, ...)
#     temp_cumulative_df = DFForCumulativePlot(temp_stack_df, window_length=window_length, dt=0.15)
#     p = p + geom_step(data=temp_cumulative_df[temp_cumulative_df$bird == "adult", ], aes(t, y), alpha=0.4, color="indianred2")
#   }
#   p
# }
# 
# #---- functions for the circular shifts plots  END  ----