def storeCalibrationParams(summary, xOff, yOff, zOff, xSlope, ySlope, zSlope,
xTemp, yTemp, zTemp, meanTemp):
"""Store calibration parameters to output summary dictionary
:param dict summary: Output dictionary containing all summary metrics
:param float xOff: x intercept
:param float yOff: y intercept
:param float zOff: z intercept
:param float xSlope: x slope
:param float ySlope: y slope
:param float zSlope: z slope
:param float xTemp: x temperature
:param float yTemp: y temperature
:param float zTemp: z temperature
:param float meanTemp: Calibration mean temperature in file
:return: Calibration summary values written to dict <summary>
:rtype: void
"""
# store output to summary dictionary
summary['calibration-xOffset(g)'] = accUtils.formatNum(xOff, 4)
summary['calibration-yOffset(g)'] = accUtils.formatNum(yOff, 4)
summary['calibration-zOffset(g)'] = accUtils.formatNum(zOff, 4)
summary['calibration-xSlope(g)'] = accUtils.formatNum(xSlope, 4)
summary['calibration-ySlope(g)'] = accUtils.formatNum(ySlope, 4)
summary['calibration-zSlope(g)'] = accUtils.formatNum(zSlope, 4)
summary['calibration-xTemp(C)'] = accUtils.formatNum(xTemp, 4)
summary['calibration-yTemp(C)'] = accUtils.formatNum(yTemp, 4)
summary['calibration-zTemp(C)'] = accUtils.formatNum(zTemp, 4)
summary['calibration-meanDeviceTemp(C)'] = accUtils.formatNum(meanTemp, 2)
def calculateECDF(e, inputCol, summary):
"""Calculate activity intensity empirical cumulative distribution
The input data must not be imputed, as ECDF requires different imputation
where nan/non-wear data segments are IMPUTED FOR EACH INTENSITY LEVEL. Here,
the average of similar time-of-day values is imputed with one minute
granularity on different days of the measurement. Following intensity levels
are calculated
1mg bins from 1-20mg
5mg bins from 25-100mg
25mg bins from 125-500mg
100mg bins from 500-2000mg
:param pandas.DataFrame e: Pandas dataframe of epoch data
:param str inputCol: Column to calculate intensity distribution on
:param dict summary: Output dictionary containing all summary metrics
:return: Write dict <summary> keys '<inputCol>-ecdf-<level...>mg'
:rtype: void
"""
ecdf1, step = np.linspace(1, 20, 20, retstep=True) # 1mg bins from 1-20mg
ecdf2, step = np.linspace(25, 100, 16, retstep=True) # 5mg bins from 25-100mg
ecdf3, step = np.linspace(125, 500, 16, retstep=True) # 25mg bins from 125-500mg
ecdf4, step = np.linspace(600, 2000, 15, retstep=True) # 100mg bins from 500-2000mg
ecdfXVals = np.concatenate([ecdf1, ecdf2, ecdf3, ecdf4])
# remove NaNs (necessary for statsmodels.api)
ecdfData = e[['hour', 'minute', inputCol]][~np.isnan(e[inputCol])]
if len(ecdfData) > 0:
# set column names for actual, imputed, and adjusted intensity dist. vals
cols = []
colsImputed = []
colsAdjusted = []
for xVal in ecdfXVals:
col = 'ecdf' + str(xVal)
cols.append(col)
colsImputed.append(col + 'Imputed')
colsAdjusted.append(col + 'Adjusted')
ecdfData[col] = (ecdfData[inputCol] <= xVal) * 1.0
# calculate imputation values to replace nan metric values
wearTimeWeights = ecdfData.groupby(['hour', 'minute'])[cols].mean()
ecdfData = ecdfData.join(wearTimeWeights, on=['hour', 'minute'],
rsuffix='Imputed')
# for each ecdf xVal column, apply missing data imputation
for col, imputed, adjusted in zip(cols, colsImputed, colsAdjusted):
ecdfData[adjusted] = ecdfData[col].fillna(ecdfData[imputed])
accEcdf = ecdfData[colsAdjusted].mean()
else:
accEcdf = pd.Series(data=[0.0 for i in ecdfXVals],
index=[str(i)+'Adjusted' for i in ecdfXVals])
# and write to summary dict
for x, ecdf in zip(ecdfXVals, accEcdf):
summary[inputCol + '-ecdf-' + str(accUtils.formatNum(x,0)) + 'mg'] = \
accUtils.formatNum(ecdf, 4)
def get_interrupts(e, epochPeriod, summary):
"""Identify if there are interrupts in the data recording
:param pandas.DataFrame e: Pandas dataframe of epoch data
:param int epochPeriod: Size of epoch time window (in seconds)
:param dict summary: Output dictionary containing all summary metrics
:return: Write dict <summary> keys 'err-interrupts-num' & 'errs-interrupt-mins'
:rtype: void
"""
e.dropna(subset=['enmoTrunc', 'xStd', 'yStd', 'zStd'],
how='all',
inplace=True)
epochNs = epochPeriod * np.timedelta64(1, 's')
interrupts = np.where(np.diff(np.array(e.index)) > epochNs)[0]
# get duration of each interrupt in minutes
interruptMins = []
for i in interrupts:
interruptMins.append(
np.diff(np.array(e[i:i + 2].index)) / np.timedelta64(1, 'm'))
# record to output summary
summary['errs-interrupts-num'] = len(interruptMins)
summary['errs-interrupt-mins'] = accUtils.formatNum(
np.sum(interruptMins), 1)
def get_interrupts(e, epochPeriod, summary):
"""Identify if there are interrupts in the data recording
:param pandas.DataFrame e: Pandas dataframe of epoch data
:param int epochPeriod: Size of epoch time window (in seconds)
:param dict summary: Output dictionary containing all summary metrics
:return: Write dict <summary> keys 'err-interrupts-num' & 'errs-interrupt-mins'
:rtype: void
"""
epochNs = epochPeriod * np.timedelta64(1, 's')
interrupts = np.where(e.index.to_series(keep_tz=True).diff() > epochNs)[0]
# Get duration of each interrupt in minutes
interruptMins = []
for i in interrupts:
interruptMins.append(e.index[i-1:i+1].to_series(keep_tz=True).diff() /
np.timedelta64(1, 'm'))
# Record to output summary
summary['errs-interrupts-num'] = len(interruptMins)
summary['errs-interrupt-mins'] = accUtils.formatNum(np.sum(interruptMins), 1)
frames = [e]
for i in interrupts:
start, end = e.index[i-1:i+1]
dti = pd.date_range(start=start, end=end, freq=str(epochPeriod)+'s')[1:-1]
frames.append(dti.to_frame().drop(columns=0))
e = pd.concat(frames).sort_index()
return e
def get_interrupts(e, epochPeriod, summary):
"""Identify if there are interrupts in the data recording
:param pandas.DataFrame e: Pandas dataframe of epoch data
:param int epochPeriod: Size of epoch time window (in seconds)
:param dict summary: Output dictionary containing all summary metrics
:return: Write dict <summary> keys 'err-interrupts-num' & 'errs-interrupt-mins'
:rtype: void
"""
epochNs = epochPeriod * np.timedelta64(1, 's')
interrupts = np.where(np.diff(np.array(e.index)) > epochNs)[0]
# get duration of each interrupt in minutes
interruptMins = []
for i in interrupts:
interruptMins.append(
np.diff(np.array(e[i:i + 2].index)) / np.timedelta64(1, 'm'))
# record to output summary
summary['errs-interrupts-num'] = len(interruptMins)
summary['errs-interrupt-mins'] = accUtils.formatNum(
np.sum(interruptMins), 1)
for i in interrupts:
start, end = e[i:i + 2].index
dti = pd.date_range(start=start, end=end,
freq=str(epochPeriod) + 's')[1:-1]
e = e.append(dti.to_frame())
e = e.sort_index()
return e
def writeMovementSummaries(e, labels, summary, useRecommendedImputation):
"""Write overall summary stats for each activity type to summary dict
:param pandas.DataFrame e: Pandas dataframe of epoch data
:param list(str) labels: Activity state labels
:param dict summary: Output dictionary containing all summary metrics
:param bool useRecommendedImputation: Highly recommended method to impute
missing data using data from other days around the same time
:return: Write dict <summary> keys for each activity type 'overall-<avg/sd>',
'week<day/end>-avg', '<day..>-avg', 'hourOfDay-<hr..>-avg',
'hourOfWeek<day/end>-<hr..>-avg'
:rtype: void
"""
# Identify activity types to summarise
activityTypes = ['acc', 'CutPointMVPA', 'CutPointVPA']
activityTypes += labels
if 'MET' in e.columns:
activityTypes.append('MET')
# Sumarise each type by: overall, week day/end, day, and hour of day
for accType in activityTypes:
col = accType
if useRecommendedImputation:
col += 'Imputed'
if accType in ['CutPointMVPA', 'CutPointVPA']:
col = accType
# Overall / weekday / weekend summaries
summary[accType + '-overall-avg'] = accUtils.formatNum(
e[col].mean(), 5)
summary[accType + '-overall-sd'] = accUtils.formatNum(e[col].std(), 2)
summary[accType + '-weekday-avg'] = accUtils.formatNum(
e[col][e.index.weekday <= 4].mean(), 2)
summary[accType + '-weekend-avg'] = accUtils.formatNum(
e[col][e.index.weekday >= 5].mean(), 2)
# Daily summary
for i, day in zip(range(0, 7), accUtils.DAYS):
summary[accType + '-' + day + '-avg'] = accUtils.formatNum(
e[col][e.index.weekday == i].mean(), 2)
# Hourly summaries
for i in range(0, 24):
hourOfDay = accUtils.formatNum(e[col][e.index.hour == i].mean(), 2)
hourOfWeekday = accUtils.formatNum(
e[col][(e.index.weekday <= 4) & (e.index.hour == i)].mean(), 2)
hourOfWeekend = accUtils.formatNum(
e[col][(e.index.weekday >= 5) & (e.index.hour == i)].mean(), 2)
# Write derived hourly values to summary dictionary
summary[accType + '-hourOfDay-' + str(i) + '-avg'] = hourOfDay
summary[accType + '-hourOfWeekday-' + str(i) +
'-avg'] = hourOfWeekday
summary[accType + '-hourOfWeekend-' + str(i) +
'-avg'] = hourOfWeekend
def storeCalibrationInformation(summary, bestIntercept, bestSlope,
bestTemp, meanTemp, initError, bestError, xMin, xMax, yMin, yMax, zMin,
zMax, nStatic, calibrationSphereCriteria = 0.3):
"""Store calibration information to output summary dictionary
:param dict summary: Output dictionary containing all summary metrics
:param list(float) bestIntercept: Best x/y/z intercept values
:param list(float) bestSlope: Best x/y/z slope values
:param list(float) bestTemperature: Best x/y/z temperature values
:param float meanTemp: Calibration mean temperature in file
:param float initError: Root mean square error (in mg) before calibration
:param float initError: Root mean square error (in mg) after calibration
:param float xMin: xMin information on spread of stationary points
:param float xMax: xMax information on spread of stationary points
:param float yMin: yMin information on spread of stationary points
:param float yMax: yMax information on spread of stationary points
:param float zMin: zMin information on spread of stationary points
:param float zMax: zMax information on spread of stationary points
:param int nStatic: number of stationary points used for calibration
:param float calibrationSphereCriteria: Threshold to check how well file was
calibrated
:return: Calibration summary values written to dict <summary>
:rtype: void
"""
# store output to summary dictionary
summary['calibration-errsBefore(mg)'] = accUtils.formatNum(initError*1000, 2)
summary['calibration-errsAfter(mg)'] = accUtils.formatNum(bestError*1000, 2)
storeCalibrationParams(summary, bestIntercept[0], bestIntercept[1],
bestIntercept[2], bestSlope[0], bestSlope[1], bestSlope[2],
bestTemp[0], bestTemp[1], bestTemp[2], meanTemp)
summary['calibration-numStaticPoints'] = nStatic
summary['calibration-staticXmin(g)'] = accUtils.formatNum(xMin, 2)
summary['calibration-staticXmax(g)'] = accUtils.formatNum(xMax, 2)
summary['calibration-staticYmin(g)'] = accUtils.formatNum(yMin, 2)
summary['calibration-staticYmax(g)'] = accUtils.formatNum(yMax, 2)
summary['calibration-staticZmin(g)'] = accUtils.formatNum(zMin, 2)
summary['calibration-staticZmax(g)'] = accUtils.formatNum(zMax, 2)
# check how well calibrated file was
summary['quality-calibratedOnOwnData'] = 1
summary['quality-goodCalibration'] = 1
s = calibrationSphereCriteria
try:
if xMin > -s or xMax < s or yMin > -s or yMax < s or zMin > -s or zMax < s or \
np.isnan(xMin) or np.isnan(yMin) or np.isnan(zMin):
summary['quality-goodCalibration'] = 0
except UnboundLocalError:
summary['quality-goodCalibration'] = 0
def writeMovementSummaries(e, labels, summary):
"""Write overall summary stats for each activity type to summary dict
:param pandas.DataFrame e: Pandas dataframe of epoch data
:param list(str) labels: Activity state labels
:param dict summary: Output dictionary containing all summary metrics
:return: Write dict <summary> keys for each activity type 'overall-<avg/sd>',
'week<day/end>-avg', '<day..>-avg', 'hourOfDay-<hr..>-avg',
'hourOfWeek<day/end>-<hr..>-avg'
:rtype: void
"""
# identify activity types to summarise
activityTypes = ['acc', 'MVPA', 'VPA']
activityTypes += labels
if 'MET' in e.columns:
activityTypes.append('MET')
# sumarise each type by: overall, week day/end, day, and hour of day
for accType in activityTypes:
col = accType + 'Imputed'
if accType in ['MVPA', 'VPA']:
col = accType
# overall / weekday / weekend summaries
summary[accType + '-overall-avg'] = accUtils.formatNum(e[col].mean(), 2)
summary[accType + '-overall-sd'] = accUtils.formatNum(e[col].std(), 2)
summary[accType + '-weekday-avg'] = accUtils.formatNum( \
e[col][e.index.weekday<=4].mean(), 2)
summary[accType + '-weekend-avg'] = accUtils.formatNum( \
e[col][e.index.weekday>=5].mean(), 2)
# daily summary
for i, day in zip(range(0, 7), accUtils.DAYS):
summary[accType + '-' + day + '-avg'] = accUtils.formatNum( \
e[col][e.index.weekday == i].mean(), 2)
# hourly summaries
for i in range(0, 24):
hourOfDay = accUtils.formatNum(e[col][e.index.hour == i].mean(), 2)
hourOfWeekday = accUtils.formatNum( \
e[col][(e.index.weekday<=4) & (e.index.hour == i)].mean(), 2)
hourOfWeekend = accUtils.formatNum( \
e[col][(e.index.weekday>=5) & (e.index.hour == i)].mean(), 2)
# write derived hourly values to summary dictionary
summary[accType + '-hourOfDay-' + str(i) + '-avg'] = hourOfDay
summary[accType + '-hourOfWeekday-' + str(i) + '-avg'] = hourOfWeekday
summary[accType + '-hourOfWeekend-' + str(i) + '-avg'] = hourOfWeekend
def storeCalibrationParams(summary, xyzOff, xyzSlope, xyzTemp, meanTemp):
"""Store calibration parameters to output summary dictionary
:param dict summary: Output dictionary containing all summary metrics
:param list(float) xyzOff: intercept [x, y, z]
:param list(float) xyzSlope: slope [x, y, z]
:param list(float) xyzTemp: temperature [x, y, z]
:param float meanTemp: Calibration mean temperature in file
:return: Calibration summary values written to dict <summary>
:rtype: void
"""
# store output to summary dictionary
summary['calibration-xOffset(g)'] = accUtils.formatNum(xyzOff[0], 4)
summary['calibration-yOffset(g)'] = accUtils.formatNum(xyzOff[1], 4)
summary['calibration-zOffset(g)'] = accUtils.formatNum(xyzOff[2], 4)
summary['calibration-xSlope(g)'] = accUtils.formatNum(xyzSlope[0], 4)
summary['calibration-ySlope(g)'] = accUtils.formatNum(xyzSlope[1], 4)
summary['calibration-zSlope(g)'] = accUtils.formatNum(xyzSlope[2], 4)
summary['calibration-xTemp(C)'] = accUtils.formatNum(xyzTemp[0], 4)
summary['calibration-yTemp(C)'] = accUtils.formatNum(xyzTemp[1], 4)
summary['calibration-zTemp(C)'] = accUtils.formatNum(xyzTemp[2], 4)
summary['calibration-meanDeviceTemp(C)'] = accUtils.formatNum(meanTemp, 2)
def get_interrupts(e, epochPeriod, summary):
"""Identify if there are interrupts in the data recording
:param pandas.DataFrame e: Pandas dataframe of epoch data
:param int epochPeriod: Size of epoch time window (in seconds)
:param dict summary: Output dictionary containing all summary metrics
:return: Write dict <summary> keys 'err-interrupts-num' & 'errs-interrupt-mins'
:rtype: void
"""
e.dropna(subset=['enmoTrunc', 'xStd', 'yStd', 'zStd'], how='all',
inplace=True)
epochNs = epochPeriod * np.timedelta64(1, 's')
interrupts = np.where(np.diff(np.array(e.index)) > epochNs)[0]
# get duration of each interrupt in minutes
interruptMins = []
for i in interrupts:
interruptMins.append( np.diff(np.array(e[i:i+2].index)) /
np.timedelta64(1, 'm') )
# record to output summary
summary['errs-interrupts-num'] = len(interruptMins)
summary['errs-interrupt-mins'] = accUtils.formatNum(np.sum(interruptMins), 1)
def calculateECDF(e, inputCol, summary):
"""Calculate activity intensity empirical cumulative distribution
The input data must not be imputed, as ECDF requires different imputation
where nan/non-wear data segments are IMPUTED FOR EACH INTENSITY LEVEL. Here,
the average of similar time-of-day values is imputed with one minute
granularity on different days of the measurement. Following intensity levels
are calculated
1mg bins from 1-20mg
5mg bins from 25-100mg
25mg bins from 125-500mg
100mg bins from 500-2000mg
:param pandas.DataFrame e: Pandas dataframe of epoch data
:param str inputCol: Column to calculate intensity distribution on
:param dict summary: Output dictionary containing all summary metrics
:return: Write dict <summary> keys '<inputCol>-ecdf-<level...>mg'
:rtype: void
"""
ecdf1, step = np.linspace(1, 20, 20, retstep=True) # 1mg bins from 1-20mg
ecdf2, step = np.linspace(25, 100, 16,
retstep=True) # 5mg bins from 25-100mg
ecdf3, step = np.linspace(125, 500, 16,
retstep=True) # 25mg bins from 125-500mg
ecdf4, step = np.linspace(600, 2000, 15,
retstep=True) # 100mg bins from 500-2000mg
ecdfXVals = np.concatenate([ecdf1, ecdf2, ecdf3, ecdf4])
# remove NaNs (necessary for statsmodels.api)
ecdfData = e[['hour', 'minute', inputCol]][~np.isnan(e[inputCol])]
if len(ecdfData) > 0:
# set column names for actual, imputed, and adjusted intensity dist. vals
cols = []
colsImputed = []
colsAdjusted = []
for xVal in ecdfXVals:
col = 'ecdf' + str(xVal)
cols.append(col)
colsImputed.append(col + 'Imputed')
colsAdjusted.append(col + 'Adjusted')
ecdfData[col] = (ecdfData[inputCol] <= xVal) * 1.0
# calculate imputation values to replace nan metric values
wearTimeWeights = ecdfData.groupby(['hour', 'minute'])[cols].mean()
ecdfData = ecdfData.join(wearTimeWeights,
on=['hour', 'minute'],
rsuffix='Imputed')
# for each ecdf xVal column, apply missing data imputation
for col, imputed, adjusted in zip(cols, colsImputed, colsAdjusted):
ecdfData[adjusted] = ecdfData[col].fillna(ecdfData[imputed])
accEcdf = ecdfData[colsAdjusted].mean()
else:
accEcdf = pd.Series(data=[0.0 for i in ecdfXVals],
index=[str(i) + 'Adjusted' for i in ecdfXVals])
# and write to summary dict
for x, ecdf in zip(ecdfXVals, accEcdf):
summary[inputCol + '-ecdf-' + str(accUtils.formatNum(x,0)) + 'mg'] = \
accUtils.formatNum(ecdf, 4)
def get_wear_time_stats(e, epochPeriod, maxStd, minDuration, nonWearFile,
summary):
"""Calculate nonWear time, write episodes to file, and return wear statistics
If daylight savings crossover, update times after time-change by +/- 1hr.
Also, if Autumn crossover time, remove last 1hr chunk before time-change.
:param pandas.DataFrame e: Pandas dataframe of epoch data
:param int epochPeriod: Size of epoch time window (in seconds)
:param int maxStd: Threshold (in mg units) for stationary vs not
:param int minDuration: Minimum duration of nonwear events (minutes)
:param str nonWearFile: Output filename for non wear .csv.gz episodes
:param dict summary: Output dictionary containing all summary metrics
:return: Write dict <summary> keys 'wearTime-numNonWearEpisodes(>1hr)',
'wearTime-overall(days)', 'nonWearTime-overall(days)', 'wearTime-diurnalHrs',
'wearTime-diurnalMins', 'quality-goodWearTime', 'wearTime-<day...>', and
'wearTime-hourOfDay-<hr...>'
:rtype: void
:return: Write .csv.gz non wear episodes file to <nonWearFile>
:rtype: void
"""
maxStd = maxStd / 1000.0 # java uses Gravity units (not mg)
e['nw'] = np.where(
(e['xStd'] < maxStd) & (e['yStd'] < maxStd) & (e['zStd'] < maxStd), 1,
0)
starts = e.index[(e['nw'] == True)
& (e['nw'].shift(1).fillna(False) == False)]
ends = e.index[(e['nw'] == True)
& (e['nw'].shift(-1).fillna(False) == False)]
nonWearEpisodes = [(start, end) for start, end in zip(starts, ends)
if end > start + np.timedelta64(minDuration, 'm')]
# set nonWear data to nan and record to nonWearBouts file
f = gzip.open(nonWearFile, 'wb')
f.write('start,end,xStdMax,yStdMax,zStdMax\n'.encode())
timeFormat = '%Y-%m-%d %H:%M:%S'
for episode in nonWearEpisodes:
tmp = e[['xStd', 'yStd', 'zStd']][episode[0]:episode[1]]
nonWearBout = episode[0].strftime(timeFormat) + ','
nonWearBout += episode[1].strftime(timeFormat) + ','
nonWearBout += str(tmp['xStd'].mean()) + ','
nonWearBout += str(tmp['yStd'].mean()) + ','
nonWearBout += str(tmp['zStd'].mean()) + '\n'
f.write(nonWearBout.encode())
# set main dataframe values to nan
e[episode[0]:episode[1]] = np.nan
f.close()
# write to summary
summary['wearTime-numNonWearEpisodes(>1hr)'] = int(len(nonWearEpisodes))
# calculate wear statistics
wearSamples = e['enmoTrunc'].count()
nonWearSamples = len(e[np.isnan(e['enmoTrunc'])].index.values)
wearTimeMin = wearSamples * epochPeriod / 60.0
nonWearTimeMin = nonWearSamples * epochPeriod / 60.0
# write to summary
summary['wearTime-overall(days)'] = accUtils.formatNum(
wearTimeMin / 1440.0, 2)
summary['nonWearTime-overall(days)'] = accUtils.formatNum(
nonWearTimeMin / 1440.0, 2)
# get wear time in each of 24 hours across week
epochsInMin = 60.0 / epochPeriod
for i, day in zip(range(0, 7), accUtils.DAYS):
dayWear = e['enmoTrunc'][e.index.weekday == i].count() / epochsInMin
# write to summary
summary['wearTime-' + day + '(hrs)'] = accUtils.formatNum(
dayWear / 60.0, 2)
for i in range(0, 24):
hourWear = e['enmoTrunc'][e.index.hour == i].count() / epochsInMin
# write to summary
summary['wearTime-hourOfDay' + str(i) + '-(hrs)'] = \
accUtils.formatNum(hourWear/60.0, 2)
summary['wearTime-diurnalHrs'] = accUtils.formatNum( \
e['enmoTrunc'].groupby(e.index.hour).mean().count(), 2)
summary['wearTime-diurnalMins'] = accUtils.formatNum( \
e['enmoTrunc'].groupby([e.index.hour, e.index.minute]).mean().count(), 2)
# write binary decision on whether weartime was good or not
minDiurnalHrs = 24
minWearDays = 3
summary['quality-goodWearTime'] = 1
if summary['wearTime-diurnalHrs'] < minDiurnalHrs or \
summary['wearTime-overall(days)'] < minWearDays:
summary['quality-goodWearTime'] = 0
def get_wear_time_stats(e, epochPeriod, maxStd, minDuration, nonWearFile,
summary):
"""Calculate nonWear time, write episodes to file, and return wear statistics
If daylight savings crossover, update times after time-change by +/- 1hr.
Also, if Autumn crossover time, remove last 1hr chunk before time-change.
:param pandas.DataFrame e: Pandas dataframe of epoch data
:param int epochPeriod: Size of epoch time window (in seconds)
:param int maxStd: Threshold (in mg units) for stationary vs not
:param int minDuration: Minimum duration of nonwear events (minutes)
:param str nonWearFile: Output filename for non wear .csv.gz episodes
:param dict summary: Output dictionary containing all summary metrics
:return: Write dict <summary> keys 'wearTime-numNonWearEpisodes(>1hr)',
'wearTime-overall(days)', 'nonWearTime-overall(days)', 'wearTime-diurnalHrs',
'wearTime-diurnalMins', 'quality-goodWearTime', 'wearTime-<day...>', and
'wearTime-hourOfDay-<hr...>'
:rtype: void
:return: Write .csv.gz non wear episodes file to <nonWearFile>
:rtype: void
"""
maxStd = maxStd / 1000.0 # java uses Gravity units (not mg)
e['nw'] = np.where((e['xStd']<maxStd) & (e['yStd']<maxStd) &
(e['zStd']<maxStd), 1, 0)
starts = e.index[(e['nw']==True) & (e['nw'].shift(1).fillna(False)==False)]
ends = e.index[(e['nw']==True) & (e['nw'].shift(-1).fillna(False)==False)]
nonWearEpisodes = [(start, end) for start, end in zip(starts, ends)
if end > start + np.timedelta64(minDuration,'m')]
# set nonWear data to nan and record to nonWearBouts file
f = gzip.open(nonWearFile,'wb')
f.write('start,end,xStdMax,yStdMax,zStdMax\n'.encode())
timeFormat = '%Y-%m-%d %H:%M:%S'
for episode in nonWearEpisodes:
tmp = e[['xStd','yStd','zStd']][episode[0]:episode[1]]
nonWearBout = episode[0].strftime(timeFormat) + ','
nonWearBout += episode[1].strftime(timeFormat) + ','
nonWearBout += str(tmp['xStd'].mean()) + ','
nonWearBout += str(tmp['yStd'].mean()) + ','
nonWearBout += str(tmp['zStd'].mean()) + '\n'
f.write(nonWearBout.encode())
# set main dataframe values to nan
e[episode[0]:episode[1]] = np.nan
f.close()
# write to summary
summary['wearTime-numNonWearEpisodes(>1hr)'] = int(len(nonWearEpisodes))
# calculate wear statistics
wearSamples = e['enmoTrunc'].count()
nonWearSamples = len(e[np.isnan(e['enmoTrunc'])].index.values)
wearTimeMin = wearSamples * epochPeriod / 60.0
nonWearTimeMin = nonWearSamples * epochPeriod / 60.0
# write to summary
summary['wearTime-overall(days)'] = accUtils.formatNum(wearTimeMin/1440.0, 2)
summary['nonWearTime-overall(days)'] = accUtils.formatNum(nonWearTimeMin/1440.0, 2)
# get wear time in each of 24 hours across week
epochsInMin = 60.0 / epochPeriod
for i, day in zip(range(0, 7), accUtils.DAYS):
dayWear = e['enmoTrunc'][e.index.weekday == i].count() / epochsInMin
# write to summary
summary['wearTime-' + day + '(hrs)'] = accUtils.formatNum(dayWear/60.0, 2)
for i in range(0, 24):
hourWear = e['enmoTrunc'][e.index.hour == i].count() / epochsInMin
# write to summary
summary['wearTime-hourOfDay' + str(i) + '-(hrs)'] = \
accUtils.formatNum(hourWear/60.0, 2)
summary['wearTime-diurnalHrs'] = accUtils.formatNum( \
e['enmoTrunc'].groupby(e.index.hour).mean().count(), 2)
summary['wearTime-diurnalMins'] = accUtils.formatNum( \
e['enmoTrunc'].groupby([e.index.hour, e.index.minute]).mean().count(), 2)
# write binary decision on whether weartime was good or not
minDiurnalHrs = 24
minWearDays = 3
summary['quality-goodWearTime'] = 1
if summary['wearTime-diurnalHrs'] < minDiurnalHrs or \
summary['wearTime-overall(days)'] < minWearDays:
summary['quality-goodWearTime'] = 0