Example output.csv

eyetracker-stats.py

import pandas as pd
import numpy as np
from itertools import groupby

########## Part 1: Import data

# Filename
fn = "DigitalScreenTestDataFiltered.tsv"

# Read file in with tab (\t) delimiter
df = pd.read_csv(fn, sep="\t")

# All AOI columns start with this phrase
AOI_prefix = 'AOI hit'

# Extract the AOI name from a column 
def AOI_from_column_name(col):
    # 1. Starts with `AOI hit [abcdefg.png - AOI Name]`
    
    # 2. Splits on the ` - `, turns into `AOI hit [abcdefg.png` and `AOI name]`
    snapshot_file, name = col.split(" - ")
    
    # 3. Remove the trailing bracket
    return name.replace("]", "")

# Create a mapping of column name to AOI name
AOIs = {col:AOI_from_column_name(col) for col in df.columns if col.startswith(AOI_prefix)}

######### Part 2: Aggregate and summarize data
        
# Create time delta column
# df['timestamp_delta'] = pd.to_timedelta(df['Recording timestamp'], unit='ms')

# Now, extract the differences in milliseconds.
# df['timestamp_delta_ms'] = df.apply(lambda row: row['timestamp_delta'].milliseconds, axis=1)
df['timestamp_delta_ms'] = ( df['Recording timestamp'] - df['Recording timestamp'].shift()).fillna(0)

# Calculate cumulative fixation time for each group of fixations.
for AOI in AOIs:
    accumulator = []
    last = 0
    for row in df.iterrows():
        current_delta = row[1]['timestamp_delta_ms'] + last
        is_fixation = bool(row[1][AOI])
    
        # Reset counter if this is not a fixation
        if not is_fixation:
            last = 0
            accumulator.append(np.nan)
        else:
            accumulator.append(current_delta)
            last = current_delta

    df['AOI_Cumulative_%s' % AOIs[AOI]] = accumulator
    
######### Part 3: Calculate statistics
data = df.filter(regex="AOI_Cumulative").to_dict()
fixations = {key:[] for key in AOIs.values()}

for AOI in AOIs.values():
    key = 'AOI_Cumulative_%s' % AOI
    grouped = groupby(data[key].values(), lambda item: (not np.isnan(item)))
    for is_fixation, values in grouped:
        if is_fixation:
            fixations[AOI].append(max(values))
            
# Export to file
stats = pd.DataFrame.from_dict(fixations, orient='index').transpose()
stats.to_csv('Stats_For_%s.csv' % fn.split(".")[0] )
print stats.describe()
print "%s AOIs, total recording time %s ms" % (len(AOIs), df['Recording timestamp'].max())