I have a dataset of several game reviews and I want to calculate the respective overall score each game had until the respective day - so basically the overall score a user saw on each given day.
The reviews are binary so it's just a vote up/down system where each 1 in the column positive marks an upvote:
game_id created positive
123 2018-07-18 1
123 2018-07-18 0
123 2018-07-18 1
123 2018-07-19 1
456 2018-06-23 1
456 2018-06-25 1
456 2018-06-25 0
456 2018-06-26 1
789 2018-07-18 1
calculating the overall mean per day is easy with
group_by(game_id, created) %>%
summarise(sum(positive)/n()
but I'm struggling with how to include the reviews of the days before.
I want it to look like this:
game_id created total_score
123 2018-07-18 0.66
123 2018-07-19 0.75
456 2018-06-23 1.0
456 2018-06-25 0.5
456 2018-06-26 0.66
789 2018-07-18 1
I thought about using combination of a loop and an if statement but am not really able to formulate it (and doubtful about its efficiency for larger datasets...)
Here's a way to achieve it using dplyr. The key here is to create an intermediate computation of cumulative sums and then use those for the ratio:
df %>%
group_by(game_id, created) %>%
summarise(pos=sum(positive), tot=n()) %>%
group_by(game_id) %>%
mutate(pct = cumsum(pos) / cumsum(tot))
# A tibble: 6 x 5
# Groups: game_id [3]
game_id created pos tot pct
<int> <fct> <int> <int> <dbl>
1 123 2018-07-18 2 3 0.667
2 123 2018-07-19 1 1 0.75
3 456 2018-06-23 1 1 1
4 456 2018-06-25 1 2 0.667
5 456 2018-06-26 1 1 0.75
6 789 2018-07-18 1 1 1
Assuming your dataframe is named df you can:
library(dplyr)
df= arrange(df, game_id,created) ## sort dataset
df$csum <- ave(df$positive, df$game_id, FUN=cumsum) ## create cumulative sum
to create the cumulative sum up for each game_id. Make sure your dataframe is sorted by game_id and created
Related
I have an event log data. For reproducible example, let's use the data from eventdataR
eventdataR::patients
## look at patient 1 sequence
eventdataR::patients %>% dplyr::filter(patient == '1')
# A tibble: 12 x 7
handling patient employee handling_id registration_ty~ time .order
<fct> <chr> <fct> <chr> <fct> <dttm> <int>
1 Registration 1 r1 1 start 2017-01-02 11:41:53 1
2 Triage and A~ 1 r2 501 start 2017-01-02 12:40:20 2
3 Blood test 1 r3 1001 start 2017-01-05 08:59:04 3
4 MRI SCAN 1 r4 1238 start 2017-01-05 21:37:12 4
5 Discuss Resu~ 1 r6 1735 start 2017-01-07 07:57:49 5
6 Check-out 1 r7 2230 start 2017-01-09 17:09:43 6
7 Registration 1 r1 1 complete 2017-01-02 12:40:20 7
8 Triage and A~ 1 r2 501 complete 2017-01-02 22:32:25 8
9 Blood test 1 r3 1001 complete 2017-01-05 14:34:27 9
10 MRI SCAN 1 r4 1238 complete 2017-01-06 01:54:23 10
11 Discuss Resu~ 1 r6 1735 complete 2017-01-07 10:18:08 11
12 Check-out 1 r7 2230 complete 2017-01-09 19:45:45 12
In the above example, we can see the sequence of handling for patient 1 over a period of time. We can imagine that different patients would have different sequences or went through different number of sequences.
Now let's say I'm interested in a specific sequence and want to know which patients had gone through this specific sequence. How can I filter this dataset by this specific sequence so that I can get to know who these patients are?
The filter_activity_presence from edeaR library can help me with identifying the unique sequences and its frequency
patients %>% traces
# A tibble: 7 x 3
trace absolute_frequen~ relative_frequen~
<chr> <int> <dbl>
1 Registration,Triage and Assessment,X-Ray,Discuss R~ 258 0.516
2 Registration,Triage and Assessment,Blood test,MRI ~ 234 0.468
3 Registration,Triage and Assessment,Blood test,MRI ~ 2 0.004
4 Registration,Triage and Assessment,X-Ray 2 0.004
5 Registration,Triage and Assessment 2 0.004
6 Registration,Triage and Assessment,X-Ray,Discuss R~ 1 0.002
7 Registration,Triage and Assessment,Blood test 1 0.002
Let's say I'm interested in sequence from row 5, that is patients who had exclusively this sequence Registration -> Triage -> Assessment, there are 2 patients who had this sequence.
It seems the library that doesn't provide ready made function to extract this. At least from this doc page, https://www.bupar.net/subsetting.html#trace_length, it's not available.
Basically, given an exhaustive list of sequence, return all the patients who had gone through exactly this sequence.
In fact, if I can rebuild the trace and map it back to the original dataset, that should allow for a simple dplyr::filter. But this may not be ideal as well in the case if I'm interested in open ended sequence, for example, find all patients who started with Registration -> Triage and can be followed by any sequence.
Here's my long-winded attempt
# get trace for each patient
patient_trace <- as_tibble(patients) %>% group_by(patient) %>% dplyr::filter(registration_type == 'complete') %>%
summarise(trace = paste(handling, collapse = ","), n = n())
# identify the sequence trace of interest
trace_summary <- patients %>% traces
# here we want to see patients who had the sequence from row 5
res <- patients %>%
dplyr::filter(patient %in% c(patient_trace %>% dplyr::filter(trace %in% trace_summary$trace[5]) %>% .$patient)) %>%
dplyr::filter(registration_type == 'complete') %>%
arrange(patient, time)
# A tibble: 4 x 7
handling patient employee handling_id registration_ty~ time .order
<fct> <chr> <fct> <chr> <fct> <dttm> <int>
1 Registration 499 r1 499 complete 2018-05-01 22:57:38 1
2 Triage and As~ 499 r2 999 complete 2018-05-04 23:53:27 3
3 Registration 500 r1 500 complete 2018-05-02 01:28:23 2
4 Triage and As~ 500 r2 1000 complete 2018-05-05 07:16:02 4
You can filter them with dplyr :
library(dplyr)
req_sequence <- c('Registration', 'Triage and Assessment')
eventdataR::patients %>%
group_by(patient) %>%
filter(all(handling == req_sequence)) %>%
filter(registration_type == 'complete') %>%
ungroup
# handling patient employee handling_id registration_type time .order
# <fct> <chr> <fct> <chr> <fct> <dttm> <int>
#1 Registration 499 r1 499 complete 2018-05-01 22:57:38 3220
#2 Registration 500 r1 500 complete 2018-05-02 01:28:23 3221
#3 Triage and Assessment 499 r2 999 complete 2018-05-04 23:53:27 3720
#4 Triage and Assessment 500 r2 1000 complete 2018-05-05 07:16:02 3721
For this case to be sure of the output and to avoid any recycling effect we can filter registration_type == 'complete' first and also add another check of length(req_sequence) equal to number of rows for the patient id.
eventdataR::patients %>%
filter(registration_type == 'complete') %>%
group_by(patient) %>%
filter(length(req_sequence) == n() && all(handling == req_sequence)) %>%
ungroup
I had an R question concerning data wrangling. A sample data set I will include is downloadable online:
x<- read.csv("http://mgimond.github.io/ES218/Data/CO2.csv")
The datatable is shown in the attached image.
Example data table
I want to create a new column, let's say "time_since". This column would look at the "Average" column and calculate the time (in this case months) since "Average" is less than 300. So in this screenshot all are >300, so the value would be "0", but the month that eventually has a value less than 300 would then be "1" (representing 1 month since it has been one month under 300). If the following months are still under 300, this would increase according to the months that go by, but as soon as it become >300 again it will reset.
Basically it would be a function that would calculate the difference in time since a conditional statement is met, then restarts when the conditional is broken across dates.
I apologize if I worded it a bit confusing but hopefully the message comes across.
Maybe you can try :
library(dplyr)
x %>%
group_by(grp = cumsum(as.integer(Average > 300))) %>%
mutate(time_since = row_number()) %>%
ungroup -> result
Just to show you one excerpt of output where time_since > 1.
result %>% filter(grp == 61)
# Year Month Average Interpolated Trend Daily_mean grp time_since
# <int> <int> <dbl> <dbl> <dbl> <int> <int> <int>
#1 1964 1 320. 320. 320. -1 61 1
#2 1964 2 -100. 320. 320. -1 61 2
#3 1964 3 -100. 321. 320. -1 61 3
#4 1964 4 -100. 322. 319. -1 61 4
Here is a data.table approach. For this example, time_since is displaying the cumulative total of rows when the Average variable is greater than 315.
x<- read.csv("http://mgimond.github.io/ES218/Data/CO2.csv")
library(data.table)
setDT(x)
x[, ':='(time_since = seq(1:.N)), keyby = .(cumsum(Average < 315))][1:10, ]
#> Year Month Average Interpolated Trend Daily_mean time_since
#> 1: 1959 1 315.62 315.62 315.70 -1 1
#> 2: 1959 2 316.38 316.38 315.88 -1 2
#> 3: 1959 3 316.71 316.71 315.62 -1 3
#> 4: 1959 4 317.72 317.72 315.56 -1 4
#> 5: 1959 5 318.29 318.29 315.50 -1 5
#> 6: 1959 6 318.15 318.15 315.92 -1 6
#> 7: 1959 7 316.54 316.54 315.66 -1 7
#> 8: 1959 8 314.80 314.80 315.81 -1 1
#> 9: 1959 9 313.84 313.84 316.55 -1 1
#> 10: 1959 10 313.26 313.26 316.19 -1 1
Created on 2021-03-17 by the reprex package (v0.3.0)
There is a longitudinal data set in the wide format, from which I want to compute time (in years and days) between the first observation date and the last date an individual was observed. Dates are in the format yyyy-mm-dd. The data set has four observation periods with missing dates, an example is as follows
df1<-data.frame("id"=c(1:4),
"adate"=c("2011-06-18","2011-06-18","2011-04-09","2011-05-20"),
"bdate"=c("2012-06-15","2012-06-15",NA,"2012-05-23"),
"cdate"=c("2013-06-18","2013-06-18","2013-04-09",NA),
"ddate"=c("2014-06-15",NA,"2014-04-11",NA))
Here "adate" is the first date and the last date is the date an individual was last seen. To compute the time difference (lastdate-adate), I have tried using "lubridate" package, for example
lubridate::time_length(difftime(as.Date("2012-05-23"), as.Date("2011-05-20")),"years")
However, I'm challenged by the fact that the last date is not coming from one column. I'm looking for a way to automate the calculation in R. The expected output would look like
id years days
1 1 2.99 1093
2 2 2.00 731
3 3 3.01 1098
4 4 1.01 369
Years is approximated to 2 decimal places.
Another tidyverse solution can be done by converting the data to long format, removing NA dates, and getting the time difference between last and first date for each id.
library(dplyr)
library(tidyr)
library(lubridate)
df1 %>%
pivot_longer(-id) %>%
na.omit %>%
group_by(id) %>%
mutate(value = as.Date(value)) %>%
summarise(years = time_length(difftime(last(value), first(value)),"years"),
days = as.numeric(difftime(last(value), first(value))))
#> # A tibble: 4 x 3
#> id years days
#> <int> <dbl> <dbl>
#> 1 1 2.99 1093
#> 2 2 2.00 731
#> 3 3 3.01 1098
#> 4 4 1.01 369
We could use pmap
library(dplyr)
library(purrr)
library(tidyr)
df1 %>%
mutate(out = pmap(.[-1], ~ {
dates <- as.Date(na.omit(c(...)))
tibble(years = lubridate::time_length(difftime(last(dates),
first(dates)), "years"),
days = lubridate::time_length(difftime(last(dates), first(dates)), "days"))
})) %>%
unnest_wider(out)
# A tibble: 4 x 7
# id adate bdate cdate ddate years days
# <int> <chr> <chr> <chr> <chr> <dbl> <dbl>
#1 1 2011-06-18 2012-06-15 2013-06-18 2014-06-15 2.99 1093
#2 2 2011-06-18 2012-06-15 2013-06-18 <NA> 2.00 731
#3 3 2011-04-09 <NA> 2013-04-09 2014-04-11 3.01 1098
#4 4 2011-05-20 2012-05-23 <NA> <NA> 1.01 369
Probably most of the functions introduced here might be quite complex. You should try to learn them if possible. Although will provide a Base R approach:
grp <- droplevels(interaction(df[,1],row(df[-1]))) # Create a grouping:
days <- tapply(unlist(df[-1]),grp, function(x)max(x,na.rm = TRUE) - x[1]) #Get the difference
cbind(df[1],days, years = round(days/365,2)) # Create your table
id days years
1.1 1 1093 2.99
2.2 2 731 2.00
3.3 3 1098 3.01
4.4 4 369 1.01
if comfortable with other higher functions then you could do:
dat <- aggregate(adate~id,reshape(df1,list(2:ncol(df1)), dir="long"),function(x)max(x) - x[1])
transform(dat,year = round(adate/365,2))
id adate year
1 1 1093 2.99
2 2 731 2.00
3 3 1098 3.01
4 4 369 1.01
Using base R apply :
df1[-1] <- lapply(df1[-1], as.Date)
df1[c('years', 'days')] <- t(apply(df1[-1], 1, function(x) {
x <- na.omit(x)
x1 <- difftime(x[length(x)], x[1], 'days')
c(x1/365, x1)
}))
df1[c('id', 'years', 'days')]
# id years days
#1 1 2.994521 1093
#2 2 2.002740 731
#3 3 3.008219 1098
#4 4 1.010959 369
I'm new to Stackoverflow and looked at similar posts but couldn't find a solution that can capture time differences from multiple events from the same ID.
What I've got:
Time<-c('2016-10-04','2016-10-18', '2016-10-04','2016-10-18','2016-10-19','2016-10-28','2016-10-04','2016-10-19','2016-10-21','2016-10-22', '2017-01-02', '2017-03-04')
Value<-c(0,1,0,1,0,0,0,1,0,1,1,0)
StoreID<-c('a','a','b','b','c','c','d','d','a','a','d','c')
Unit<-c(1,1,2,2,5,5,6,6,1,1,6,5)
Helper<-c('a1','a1','b2','b2','c5','c5','d6','d6','a1','a1','d6','c5')
The helper column is the StoreID and Unit combined because I couldn't figure out how to group by both Store ID and the Unit. I want to sort the data to show when the unit was disabled (value =0) and enabled again (value =1).
Ultimately, I'd want:
Store_ID Unit Helper Time(v=0) Time(v=1) Time2(v=0) Time 2(v=1)
a 1 a1 2016-10-04 2016-10-18 2016-10-21 2016-10-22
b 2 b2 2016-10-04 2016-10-18
c 5 c5 2016-10-19 2016-10-28 2017-03-04
d 6 d6 2016-10-04 2017-10-19
Any thoughts?
I'm thinking something in dplyr but am stumped about where to go further.
Create a Header column that combines the Value column and the row number that distinguishes duplicates, then spread to wide format:
Didn't use the helper column, grouped by StoredID and Unit instead.
df <- data.frame(StoreID, Unit, Time, Value)
df %>%
group_by(StoreID, Unit, Value) %>%
mutate(Headers = sprintf('Time %s (v=%s)', row_number(), Value)) %>%
ungroup() %>% select(-Value) %>%
spread(Headers, Time)
# A tibble: 4 x 7
# StoreID Unit `Time 1 (v=0)` `Time 1 (v=1)` `Time 2 (v=0)` `Time 2 (v=1)` `Time 3 (v=0)`
#* <fctr> <dbl> <fctr> <fctr> <fctr> <fctr> <fctr>
#1 a 1 2016-10-04 2016-10-18 2016-10-21 2016-10-22 NA
#2 b 2 2016-10-04 2016-10-18 NA NA NA
#3 c 5 2016-10-19 NA 2016-10-28 NA 2017-03-04
#4 d 6 2016-10-04 2016-10-19 NA 2017-01-02 NA
Having the following table which comprises some key columns which are: customer ID | order ID | product ID | Quantity | Amount | Order Date.
All this data is in LONG Format, in that you will get multi line items for the 1 Customer ID.
I can get the first date last date using R DateDiff but converting the file to WIDE format using Plyr, still end up with the same problem of getting multiple orders by customer, just less rows and more columns.
Is there an R function that extends R DateDiff to work out how to get the time interval between purchases by Customer ID? That is, time between order 1 and 2, order 2 and 3, and so on assuming these orders exists.
CID Order.Date Order.DateMY Order.No_ Amount Quantity Category.Name Locality
1 26/02/13 Feb-13 zzzzz 1 r MOSMAN
1 26/05/13 May-13 qqqqq 1 x CHULLORA
1 28/05/13 May-13 wwwww 1 r MOSMAN
1 28/05/13 May-13 wwwww 1 x MOSMAN
2 19/08/13 Aug-13 wwwwww 1 o OAKLEIGH SOUTH
3 3/01/13 Jan-13 wwwwww 1 x CURRENCY CREEK
4 28/08/13 Aug-13 eeeeeee 1 t BRISBANE
4 10/09/13 Sep-13 rrrrrrrrr 1 y BRISBANE
4 25/09/13 Sep-13 tttttttt 2 e BRISBANE
It is not clear what do you want to do since you don't give the expected result. But I guess you want to the the intervals between 2 orders.
library(data.table)
DT <- as.data.table(DF)
DT[, list(Order.Date,
diff = c(0,diff(sort(as.Date(Order.Date,'%d/%m/%y')))) ),CID]
CID Order.Date diff
1: 1 26/02/13 0
2: 1 26/05/13 89
3: 1 28/05/13 2
4: 1 28/05/13 0
5: 2 19/08/13 0
6: 3 3/01/13 0
7: 4 28/08/13 0
8: 4 10/09/13 13
9: 4 25/09/13 15
Split the data frame and find the intervals for each Customer ID.
df <- data.frame(customerID=as.factor(c(rep("A",3),rep("B",4))),
OrderDate=as.Date(c("2013-07-01","2013-07-02","2013-07-03","2013-06-01","2013-06-02",
"2013-06-03","2013-07-01")))
dfs <- split(df,df$customerID)
lapply(dfs,function(x){
tmp <-diff(x$OrderDate)
tmp
})
Or use plyr
library(plyr)
dfs <- dlply(df,.(customerID),function(x)return(diff(x$OrderDate)))
I know this question is very old, but I just figured out another way to do it and wanted to record it:
> library(dplyr)
> library(lubridate)
> df %>% group_by(customerID) %>%
mutate(SinceLast=(interval(ymd(lag(OrderDate)),ymd(OrderDate)))/86400)
# A tibble: 7 x 3
# Groups: customerID [2]
customerID OrderDate SinceLast
<fct> <date> <dbl>
1 A 2013-07-01 NA
2 A 2013-07-02 1.
3 A 2013-07-03 1.
4 B 2013-06-01 NA
5 B 2013-06-02 1.
6 B 2013-06-03 1.
7 B 2013-07-01 28.