1. Overview
In this Exercise, we will reveal the daily routine of two randomly selected participant of the city of Engagement, Ohio USA. I will be exploring ViSIElse to visualise their routines.
2. Data Preparation
2.1 Data Packages
The following packages will be installed to provide several function for data preparation and visualisation.
- tidyverse: A package with multiple packages that provide the user functions for wrangling, visualising and modelling.
- ViSiElse: A Visual Tool for Behavior Analysis over Time.
- lubridate: A package that intuitively changes the date-time object to the user preference.
packages = c('tidyverse', 'ViSiElse', 'lubridate')
for(p in packages){
if(!require(p, character.only = T)){
install.packages(p)
}
library(p, character.only = T)
}
2.2 Importing of Dataset
Since we are observing the daily routine of the participant, data from 5 status log will be retrieved to observed the participants routine in a 1 month timeframe. To read and combine all the 5 CSVs into one data frame, we first create a function to read and merge the data. We then subseqeuntly call the function with the path of your CSV file.
multmerge = function(mypath){
filenames=list.files(path=mypath, full.names=TRUE)
datalist = lapply(filenames, function(x){read.csv(file=x,header=T)})
Reduce(function(x,y) {merge(x,y,all = TRUE)}, datalist)
}
full_data = multmerge("rawdata/")
2.3 Data Wrangling and Exploration
Retrieving the specific ID from the Data Frame
To retrieve only participant ID 1 and
300 , I first create a data frame to consist of the two
numbers. I use the semi_join function to locate the ID of 1 and 300 to
form a new data frame, subsequently using the select function to
retrieve only the timestamp, participantID,
currentMode, hungerStatus and
sleepstatus for this analysis.
FilterDF <- data.frame(participantId = c(1, 300), stringsAsFactors = FALSE)
full_data <- semi_join(full_data, FilterDF, by = "participantId")
full_data <- full_data %>%
select(c("timestamp", "participantId", "currentMode", "hungerStatus","sleepStatus"))
Splitting the Data Frame
Subsequently, I split the data frame into two different data frame based on their ID to facilitate the creation of the ViSIElse book.
participant1_data <- full_data[full_data$participantId == 1, ]
participant300_data <- full_data[full_data$participantId == 300, ]
Inserting a new column
Since the ViSIElse book require the time to be in minutes format and
an Id, using the lubridate package, I converted the
timestamp into Year-Mon-Date-Hour-Minutes-Seconds format. I then create
another column call Minutes using the mutate function and
call the hour, minute and day
function from lubridate for the creation of the two new
column Minutes and Day.
Res <- ymd_hms(participant1_data$timestamp)
participant1_data <- participant1_data %>%
mutate (Minutes = hour(Res)*60 + minute(Res), Day = day(Res))
participant1_data <- participant1_data %>%
select(c("participantId","Day","Minutes", "currentMode", "sleepStatus", "hungerStatus" )) %>%
rename ( "Current_Mode" = "currentMode",
"Sleep_Status" = "sleepStatus"
)
Generating the Final_Status
This is the tricky part, to generate the status of the participant,
we need to identify when the participant start the activity and when the
activity stop. We first construct a new column for all the required
activities using cumsum. This function will construct an
increment in value if it did not fulfil the condition inside.
From the previously created column, we will now identify the change
in status using the lead and lag function. If
the current value is different from the subsequent value (using the
lead function), there is a change in status of your
activity which will initate a stop of that activity. Similar to that, if
my current value is different from the previous value, it initate a
start of a new activity using the lag function. This will
generate the required labels to identify the change in status of the
activities.
For meal time, due to the inconsistent data provided, we will use the last hunger state as the end of meal time.
participant1_data_3 <- participant1_data_2 %>%
mutate(Sleep_label = ifelse(Sleeping != lead(Sleeping, default = 0), "Stop_Sleeping",
ifelse(Sleeping == lag(Sleeping, default = 0), "Start_Sleeping",""))) %>%
mutate(Work_label = ifelse(Working != lead(Working, default = 0), "Stop_Working",
ifelse(Working == lag(Working, default = 0), "Start_Working",""))) %>%
mutate(Recreation_label = ifelse(Recreation != lead(Recreation, default = 0), "Stop_Recreation",
ifelse(Recreation == lag(Recreation, default = 0), "Start_Recreation",""))) %>%
mutate(Hunger_label = ifelse(Hunger != lead(Hunger, default = 0), "Start_Meal", ""))
Lastly, we will now check for the change in labels to finalise the start and stop of an activity, the period of the activity will be converted to NA.
participant1_data_4 <- participant1_data_3 %>%
mutate(sleep_status = ifelse(Sleep_label == lag(Sleep_label, default = "NA"), NA , Sleep_label)) %>%
mutate(work_status = ifelse(Work_label == lag(Work_label, default = "NA"), NA , Work_label)) %>%
mutate(recreation_status = ifelse(Recreation_label == lag(Recreation_label, default = "NA"), NA, Recreation_label)) %>%
mutate(hunger_status = ifelse(Hunger_label == lag(Hunger_label, default = "NA"), NA , Hunger_label))
To identify the different meal time we will create a vector in minutes to indicate the different timing for the different meal.
supper <- 0:359
breakfast <- 360:720
lunch <- 721:1080
dinner <- 1081:1435
We will then use the case_when function to loop the
column of hunger_status to verify the different meals based
on Minutes.
participant1_data_4 <- participant1_data_4 %>%
mutate (final_meal_status = case_when(
Minutes %in% supper & (hunger_status == 'Start_Meal') & (Sleep_Status != 'Sleeping') ~ "Supper",
Minutes %in% breakfast & (hunger_status == 'Start_Meal') & (Sleep_Status != 'Sleeping') ~ "Breakfast",
Minutes %in% lunch & (hunger_status == 'Start_Meal') & (Sleep_Status != 'Sleeping') ~ "Lunch",
Minutes %in% dinner & (hunger_status == 'Start_Meal') & (Sleep_Status != 'Sleeping') ~ "Dinner")
)
Selecting the required rows for the analysis
We will force the first row of all the different status except sleep to be NA using the following function. We will then select the required columns for the VisiElse table.
participant1_data_4$work_status[1] = NA
participant1_data_4$recreation_status[1] = NA
participant1_data_4 <- participant1_data_4 %>%
select(c("Day", "Minutes", "sleep_status", "work_status", "recreation_status","final_meal_status"))
Next, we will change all blank cell to NA and subsequently unite all
the columns into one column removing all the NA value using the
unite function. For rows that have only NA, we will remove
it using the drop_na function.
participant1_data_4 [participant1_data_4 ==""] <- NA
final_data <- participant1_data_4 %>% unite ("Final_Status", sleep_status:final_meal_status, na.rm = T, remove = TRUE)
final_data [final_data ==""] <- NA
final_data <- final_data %>%
drop_na(Final_Status)
For the VisiElse package to work, the status will need to be in
columns not rows. We will use the pivot_wider function to
pivot the table with the columns as Final_Status and the
values from Minutes. Since there are some duplicates that
caused the vectors to be created, we will use the sapply
function with the head or tail to retrieve
either the first or last value of any vectors found within the columns.
For recreation, we will only find out the last recreation activity by
the participant.
pivot_table_1 <- final_data %>%
pivot_wider(names_from = Final_Status, values_from = Minutes)
pivot_table_1$Start_Working <- sapply(pivot_table_1$Start_Working, head, 1)
pivot_table_1$Stop_Working <- sapply(pivot_table_1$Stop_Working, tail, 1)
pivot_table_1$Start_Recreation <- sapply(pivot_table_1$Start_Recreation, tail, 1)
pivot_table_1$Stop_Recreation <- sapply(pivot_table_1$Stop_Recreation, tail, 1)
colnames(pivot_table_1)[1] <- "id"
Since the VisiElse package is unable to read NULL values, I will
insert the median value within the column to replace the NULL value. I
will us the as.integer() function to change all the type to
int intead of dbl.
pivot_table_1 <- readRDS("data/participants_1_clean.rds")
pivot_table_2 <- readRDS("data/participants_300_clean.rds")
x <- visielse(pivot_table_1, informer = NULL,
doplot = F,
pixel = 5)
y <- visielse(pivot_table_2, informer = NULL,
doplot = F,
pixel = 5)
x1 <- ConvertFromViSibook(x@book)
y1 <- ConvertFromViSibook(y@book)
Attempt to change the data
We will first change the order and identify which activity is long. We will then use the required variables to match the activity and recreate a book for the plot to reference. There is an error in my code despite multiple attempt to salvage it.
x1 <- x1[order(as.numeric(x1$showorder)), ]
x1$label <- c("Sleep", "Awake", "Work", "Stop Work", "Breakfast", "Have Fun", "Stop Fun", "Dinner")
x1[9,] <- c("sleep", "Sleeping", "l", 1, "Start_Sleeping", "Stop_Sleeping")
x1[10,] <- c ("work", "Working", "l", 3, "Start_Working", "Stop_Working")
x1[11,] <- c ("recreation", "Recreation Time", "l", 4, "Start_Recreation", "Stop_Recreation")
x1$showorder <- c(NA, NA, NA, NA, 2, NA , NA ,5 ,1 ,3 ,4)
x1 <- x1[order(as.numeric(x1$showorder)), ]
x1
v1 <- visielse(pivot_table_1, informer = NULL,
book = x1,
doplot = F,
pixel = 5)
plot(x,
vp0w = 0.7,
unit.tps = "min",
scal.unit.tps = 15,
main = "Typical day for Participant 1")
plot(y,
vp0w = 0.7,
unit.tps = "min",
scal.unit.tps = 15,
main = "Typical day for Participant 300")
Analysis of Results
From the analysis, we can see that Participant 1 usually goes to bed earlier but wakes up at the same time as Participant 300. Their working duration and meal time seems to be the same but we observed that Participant 3 have more night activities and sometimes all the way till midnight. This might be the age of participant 300 younger than participant 1.
Conclusion
Visielse package is not user-friendly and many errors along the way. The documentation of the package could be improved for user to replicate the data.