tests/testthat/test_correctness_dynamiccov_walkcreation.R
In bachmannpatrick/CLVTools: Tools for Customer Lifetime Value Estimation
skip_on_cran()
data("apparelTrans")
data("apparelDynCov")
test_that("interval created with eps same as if used shift()", {
skip_on_cran()
clv.time <- clv.time.weeks("ymd")
dt.interval <- pnbd_dyncov_covariate_add_interval_bounds(dt.cov=copy(apparelDynCov), clv.time=clv.time)
# same as if shifted + short manual
# no NAs in tp columns
dt.shift <- copy(apparelDynCov)
dt.shift[, tp.cov.lower := Cov.Date]
dt.shift[, tp.cov.upper := shift(tp.cov.lower, n=-1L), by="Id"]
dt.shift[, tp.cov.upper := tp.cov.upper - clv.time.epsilon(clv.time)]
dt.shift[is.na(tp.cov.upper), tp.cov.upper := tp.cov.lower + clv.time.number.timeunits.to.timeperiod(clv.time, user.number.periods=1L) - clv.time.epsilon(clv.time)]
expect_false(anyNA(dt.interval))
expect_false(anyNA(dt.shift))
expect_true(all(dt.interval == dt.shift))
})
test_that("Math for d_x is correct (Excel, tab Walk d)", {
skip_on_cran()
# From accompanying Excel, tab "Walk d"
clv.week <- clv.time.weeks("ymd")
# assuming Sunday (day 7) being the start of the week
oldval <- getOption("lubridate.week.start", 7)
options(lubridate.week.start=7)
# Monday = 2007-01-01
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-01")), 6/7)
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-02")), 5/7)
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-03")), 4/7)
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-04")), 3/7)
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-05")), 2/7)
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-06")), 1/7)
# Sunday, on boundary
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-07")), 1)
options(lubridate.week.start=oldval)
})
test_that("d_x changes correctly on (lower) boundary", {
skip_on_cran()
fct.expect.d.between.0.1 <- function(clv, d){
d <- clv.time.convert.user.input.to.timepoint(clv, user.timepoint = d)
expect_true(pnbd_dyncov_walk_d(clv.time=clv, tp.relevant.transaction = d) > 0)
expect_true(pnbd_dyncov_walk_d(clv.time=clv, tp.relevant.transaction = d) < 1)
}
fct.expect.d.equal.1 <- function(clv, d){
d <- clv.time.convert.user.input.to.timepoint(clv, user.timepoint = d)
expect_true(pnbd_dyncov_walk_d(clv.time=clv, tp.relevant.transaction = d) == 1)
}
# year
# 2006-12-31 < 1 and >0 (just before (lower) boundary)
# 2007-01-02: first after boundary, >0 and <1
# 2006-01-01: on (lower) boundary -> change -> 1
# 2007-01-01: on (lower) boundary -> 1
# 2008-01-01: on (lower) boundary -> 1 (also test for leap year)
clv.year <- clv.time.years("ymd")
fct.expect.d.between.0.1(clv.year, d="2005-12-31")
fct.expect.d.between.0.1(clv.year, d="2007-01-02")
fct.expect.d.equal.1(clv.year, d="2006-01-01")
fct.expect.d.equal.1(clv.year, d="2007-01-01")
fct.expect.d.equal.1(clv.year, d="2008-01-01")
# week
# 2007-01-01: monday, first day after (lower) boundary
# 2007-01-07: sunday, on (lower) boundary -> change -> 1
clv.week <- clv.time.weeks("ymd")
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-01")), 6/7)
fct.expect.d.equal.1(clv.week, d="2007-01-07")
# day
# ** TODO: changes always??, always == 1??
# clv.day <- clv.time.days("ymd")
# pnbd_dyncov_walk_d(clv.time=clv.day, tp.relevant.transaction = lubridate::ymd("2007-01-01"))
# hour
# 00:00:01: < 1 and > 0 (first after boundary)
# 00:00:59: < 1
# 00:01:00: < 1
# 00:59:00: < 1
# 00:59:59: < 1 and > 0 (just before boundary)
# 00:00:00: == 1
# 00:01:00: == 1
clv.hour <- clv.time.hours("ymd HMS")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:00:01")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:00:59")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:01:00")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:59:00")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:59:59")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:59:59")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 23:59:59")
# on lower boundary
fct.expect.d.equal.1(clv.hour, d="2007-01-01 00:00:00")
fct.expect.d.equal.1(clv.hour, d="2007-01-01 01:00:00")
fct.expect.d.equal.1(clv.hour, d="2007-01-01 23:00:00")
fct.expect.d.equal.1(clv.hour, d="2007-12-31 23:00:00")
})
test_that("d_omega is actually correct in cbs", {
skip_on_cran()
apparelTrans.spaced <- copy(apparelTrans)
# Fix data: space customers' first transaction out over whole first week
ids <- apparelTrans.spaced[, unique(Id)]
apparelTrans.spaced[, first.trans := Date == min(Date), by="Id"]
tp.first.trans <- apparelTrans.spaced[, min(Date)]
for(i in 1:7){
apparelTrans.spaced[Id == ids[i] & first.trans == TRUE, Date := tp.first.trans+days(i-1)]
}
clv.dyn.spaced <- fct.helper.create.clvdata.apparel.dyncov(apparelTrans.spaced, apparelDynCov, estimation.split = 38)
dt.cbs <- pnbd_dyncov_cbs(clv.dyn.spaced)
# We assume Sunday (7) to be start
dt.cbs[, weekday := lubridate::wday(date.first.actual.trans, label=TRUE, abbr=TRUE)]
dt.cbs[weekday == "Mon", unique(d_omega) == 6/7]
dt.cbs[weekday == "Tue", unique(d_omega) == 5/7]
dt.cbs[weekday == "Wed", unique(d_omega) == 4/7]
dt.cbs[weekday == "Thu", unique(d_omega) == 3/7]
dt.cbs[weekday == "Fri", unique(d_omega) == 2/7]
dt.cbs[weekday == "Sat", unique(d_omega) == 1/7]
dt.cbs[weekday == "Sun", unique(d_omega) == 1]
})
test_that("tjk is correctly calculated", {
skip_on_cran()
# Fake transactions for customer 1
# - dates: 2005-01-03, 2005-01-04, 2005-01-05, 2005-01-07, 2005-01-18
# - Id: Replace existing transactions of "1" because also requires covariates
#
# 2005-01-03 -> 2005-01-04: tjk=1/7
# 2005-01-04 -> 2005-01-05: tjk=1/7
# 2005-01-05 -> 2005-01-07: tjk=2/7
# 2005-01-07 -> 2005-01-18: tjk=11/7
apparelTrans.tjk <- rbindlist(list(apparelTrans[Id != "1"],
data.table(Id="1", Date=lubridate::ymd(c("2005-01-03", "2005-01-04", "2005-01-05", "2005-01-07", "2005-01-18")), Price=12.34)))
expect_silent(l.walks <- pnbd_dyncov_createwalks(fct.helper.create.clvdata.apparel.dyncov(apparelTrans.tjk, apparelDynCov, estimation.split = NULL)))
expect_equal(unique(l.walks$data.walks.trans.real[Id == "1", c("tp.this.trans", "tjk")])[order(tp.this.trans)],
data.table(tp.this.trans=lubridate::ymd(c("2005-01-04", "2005-01-05", "2005-01-07", "2005-01-18")),
tjk=c(1/7, 1/7, 2/7, 11/7)))
})
test_that("tjk is correctly calculated for aux trans with t.x=Tcal", {
skip_on_cran()
tp.T <- lubridate::ymd("2006-01-01")
# ensure at least one with t.x=T
apparelTrans.zeroaux <- rbindlist(list(apparelTrans, data.table(Id="1", Date=tp.T, Price=12.34)))
clv.dyn.zeroaux <- fct.helper.create.clvdata.apparel.dyncov(apparelTrans.zeroaux, apparelDynCov, estimation.split = tp.T)
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv.dyn.zeroaux))
ids.tx.eq.T <- apparelTrans.zeroaux[Date == tp.T, Id]
expect_true(l.walks$data.walks.trans.aux[Id %in% ids.tx.eq.T, all(tjk == 0)])
# Coincidence: T is also on the boundary!
expect_true(l.walks$data.walks.trans.aux[Id %in% ids.tx.eq.T, all(d1 == 1)])
})
test_that("Aux walk splitting method correct", {
# this is not from test case excel
skip_on_cran()
clv.time <- clv.time.weeks("ymd")
# Start of week: Monday = 2007-01-01
dt.cov <- data.table(Id="1",
tp.cov.lower = lubridate::ymd(c("2007-01-01", "2007-01-08", "2007-01-15")),
tp.cov.upper = lubridate::ymd(c("2007-01-07", "2007-01-14", "2007-01-21")),
cov.data = c(0.123, 0.678, 2.345))
fct.get.walk <- function(tp.last.trans, tp.estimation.end){
dt.tp.first.last <- data.table(tp.last.trans = lubridate::ymd(tp.last.trans), Id="1")
clv.time@timepoint.estimation.end <- lubridate::ymd(tp.estimation.end)
return(pnbd_dyncov_createwalks_auxwalk(dt.cov, dt.tp.first.last, names.cov="cov.data", clv.time = clv.time))
}
dt.walk <- fct.get.walk(tp.last.trans = "2007-01-05", tp.estimation.end = "2007-01-07")
expect_true(dt.walk[, .N] == 1)
expect_true(dt.walk[, cov.data] == 0.123)
dt.walk <- fct.get.walk(tp.last.trans = "2007-01-05", tp.estimation.end = "2007-01-08")
expect_true(dt.walk[, .N] == 2)
expect_true(all(dt.walk[, cov.data] == c(0.123, 0.678)))
})
test_that("Aux walk is 2 periods if T is on week start and alive at T-1 one day before (and has no real life walk)", {
# T on week start, alive is T-1 -> aux walk is 2 periods, but no real life walk (run test with T on all 7 days)
for(i in seq(7)){
data.trans <- copy(apparelTrans)
data.cov <- copy(apparelDynCov)
oldval <- getOption("lubridate.week.start")
options(lubridate.week.start=(7+i-1) %% 7+1)
# print(getOption("lubridate.week.start"))
# make cov dates to be on week start
data.cov[, Cov.Date := Cov.Date + i]
# split is on start of week
date.estimation.split <- lubridate::ymd("2005-06-26") + i
# Id==1 is zero-repeater
data.trans[Id==1, Date := date.estimation.split-1]
clv <- fct.helper.create.clvdata.apparel.dyncov(data.apparelTrans=data.trans,
data.apparelDynCov=data.cov,
estimation.split=date.estimation.split)
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv))
# has 2 aux walks
expect_true(l.walks$data.walks.life.aux[Id==1, .N] == 2)
# no real walk
expect_true(l.walks$data.walks.life.real[Id==1, .N] == 0)
options(lubridate.week.start=oldval)
}
})
test_that("Aux walks not lost if there are covariates only for the calibration period (see #134)", {
skip_on_cran()
clv.short <- fct.helper.create.clvdata.apparel.dyncov(data.apparelTrans=apparelTrans[Date <= "2005-12-31"],
data.apparelDynCov=apparelDynCov[Cov.Date <= "2005-12-31"],
estimation.split=NULL)
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv.short))
expect_true(l.walks$data.walks.life.aux[, uniqueN(Id)] == 250)
expect_true(l.walks$data.walks.trans.aux[, uniqueN(Id)] == 250)
})
test_that("Real Trans Walk correct", {
skip_on_cran()
# Fake transactions for customer "1"
# - Id: Replace existing transactions of "1" because also requires covariates
# - on and off the cov boundary
# - cov boundaries: [2005-01-02, 2005-01-08][2005-01-09, 2005-01-15][2005-01-16, 2005-01-22][2005-01-23, 2005-01-29][2005-01-30, 2005-02-05]
# - cov values: 0.123 0.234 0.345 0.456 0.567
# - transaction dates: 2005-01-02, 2005-01-07, 2005-01-10, 2005-01-15, 2005-01-16, 2005-01-30
#
# Real Trans Walk: Covariates up to a repeat transaction
# Resulting real trans walks
# 2005-01-07: [0.123]
# 2005-01-10: [0.123], [0.234]
# 2005-01-15: [0.234]
# 2005-01-16: [0.234], [0.345]
# 2005-01-30: [0.345], [0.456], [0.567]
#
apparelTrans.realwalk <- rbindlist(list(apparelTrans[Id != "1"],
data.table(Id="1", Date=lubridate::ymd(c("2005-01-02", "2005-01-07", "2005-01-10", "2005-01-15", "2005-01-16", "2005-01-30")), Price=12.34)))
dates.cov.realwalk <- lubridate::ymd(c("2005-01-02", "2005-01-09", "2005-01-16", "2005-01-23", "2005-01-30"))
values.cov.realwalk <- c(0.123, 0.234, 0.345, 0.456, 0.567)
cov.realwalk <- rbindlist(list(apparelDynCov[!(Id == "1" & Cov.Date %in% dates.cov.realwalk)],
data.table(Id="1", Cov.Date=dates.cov.realwalk, Marketing=c(0.123, 0.234, 0.345, 0.456, 0.567), Gender=0, Channel=0)))
expect_silent(l.walks <- pnbd_dyncov_createwalks(fct.helper.create.clvdata.apparel.dyncov(apparelTrans.realwalk, cov.realwalk, estimation.split = NULL)))
expect_equal(
l.walks$data.walks.trans.real[Id==1][order(tp.this.trans), list(tp.this.trans=as.character(tp.this.trans), Marketing)],
rbindlist(list(
data.table(tp.this.trans="2005-01-07", Marketing=0.123),
data.table(tp.this.trans="2005-01-10", Marketing=c(0.123, 0.234)),
data.table(tp.this.trans="2005-01-15", Marketing=0.234),
data.table(tp.this.trans="2005-01-16", Marketing=c(0.234, 0.345)),
data.table(tp.this.trans="2005-01-30", Marketing=c(0.345, 0.456, 0.567)))))
})
test_that("All walks have basic correctness, estimation.split at every day of week", {
skip_on_cran()
# From walk creation asserts function
# Test with estimation.split at every week day
for(i in seq(7)){
clv <- fct.helper.create.clvdata.apparel.dyncov(data.apparelTrans=apparelTrans,
data.apparelDynCov=apparelDynCov,
# move split by i days
estimation.split=lubridate::ymd("2005-06-30") + i)
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv))
dt.cbs <- pnbd_dyncov_cbs(clv)
fct.walk.basic.correctness <- function(dt.walks){
# no NA# coreelements
expect_true(setequal(key(dt.walks), c("Id", "walk_id", "tp.this.trans", "tp.cov.lower")))
expect_true(dt.walks[, !is.unsorted(abs_pos, strictly = TRUE)])
expect_true(dt.walks[, !is.unsorted(walk_id, strictly = FALSE)])
expect_true(dt.walks[, !is.unsorted(walk_from, strictly = FALSE)])
expect_true(dt.walks[, !is.unsorted(walk_to, strictly = FALSE)])
expect_true(!anyNA(dt.walks))
# walks are backwards looking from second transaction: transaction must be after all of walks' lower covariate periods
# all transactions must be after all of walks' lower covariate periods
}
fct.walk.basic.correctness(l.walks$data.walks.life.aux)
fct.walk.basic.correctness(l.walks$data.walks.life.real)
fct.walk.basic.correctness(l.walks$data.walks.trans.aux)
fct.walk.basic.correctness(l.walks$data.walks.trans.real)
expect_true(l.walks$data.walks.trans.real[tp.this.trans < tp.cov.lower, .N] == 0)
# All transactions on week start are longer than 1 walk (because also include cov of period before)
l.walks$data.walks.trans.real[wday(tp.this.trans) == getOption("lubridate.week.start", 1), walk_id]
# number of real trans walks == num repeat transactions
# all aux walks:
# every Id only once
# every Id in walks
# same length per customer
expect_true(l.walks$data.walks.life.aux[, list(num_walks = uniqueN(walk_id)), by="Id"][, all(num_walks==1)])
expect_true(l.walks$data.walks.life.aux[, uniqueN(Id)] == apparelTrans[, uniqueN(Id)])
expect_true(setequal(l.walks$data.walks.life.aux[, unique(Id)], apparelTrans[, unique(Id)]))
expect_true(l.walks$data.walks.trans.aux[, list(num_walks = uniqueN(walk_id)), by="Id"][, all(num_walks==1)])
expect_true(l.walks$data.walks.trans.aux[, uniqueN(Id)] == apparelTrans[, uniqueN(Id)])
expect_true(setequal(l.walks$data.walks.trans.aux[, unique(Id)], apparelTrans[, unique(Id)]))
expect_true(identical(l.walks$data.walks.life.aux[, .N, keyby="Id"],
l.walks$data.walks.trans.aux[, .N, keyby="Id"]))
# lifetime aux walk:
# no date overlap with real lifetime walks
dt.tmp <- l.walks$data.walks.life.aux[, list(first_cov_aux = min(tp.cov.lower)), keyby="Id"]
dt.tmp[l.walks$data.walks.life.real[, list(first_cov_real = min(tp.cov.lower)) , keyby="Id"], first_cov_real := i.first_cov_real, on="Id"]
# some first_cov_real are NA because have no real walk
expect_true(dt.tmp[first_cov_aux < first_cov_real, .N] == 0)
# lifetime real walk:
# exactly 1 walk per customer
# all ids except where aux walk reaches to the first transactions (coming alive)
# (number of customers = num customers in trans real walks where .N>1)
# n real walk + n aux walk >= ceiling(Tcal)
expect_true(l.walks$data.walks.life.real[, list(num_walks=uniqueN(walk_id)), keyby="Id"][, all(num_walks == 1)])
# dt.tmp[clv.fitted@clv.data@data.transactions[, list(last_trans = max(Date)), by="Id"], last_trans := i.last_trans, on="Id"]
dt.tmp[dt.cbs, first_trans := i.date.first.actual.trans, on="Id"]
expect_true(setequal(l.walks$data.walks.life.real[, unique(Id)],
dt.tmp[first_cov_aux > first_trans, Id]))
# trans real walks:
# every Id with x>0 is in ...
# ... with x-1 walks
expect_true(setequal(l.walks$data.walks.trans.real[, unique(Id)], dt.cbs[x>0, Id]))
expect_true(identical(l.walks$data.walks.trans.real[, list(num_walks=as.double(uniqueN(walk_id))), keyby="Id"],
dt.cbs[x>0, list(num_walks=as.double(x)), keyby="Id"]))
# trans walks
# tjk >= 0 (== 0 when t.x=T)
expect_true(l.walks$data.walks.trans.aux[tjk < 0, .N] == 0)
expect_true(l.walks$data.walks.trans.real[tjk < 0, .N] == 0)
# d1 and d_omega measures are in (0,1]
expect_true(dt.cbs[, all(d_omega > 0 & d_omega <= 1)])
expect_true(l.walks$data.walks.trans.aux[, all(d1 > 0 & d1 <= 1)])
expect_true(l.walks$data.walks.trans.real[, all(d1 > 0 & d1 <= 1)])
}
})
test_that("Real Trans Walk do no lose walk if transactions only 1 eps apart", {
skip_on_cran()
# real trans walk gives walk and not lost if transactions only 1 eps apart
# Customer 1 is zero repeater. Add 8 transactions only 1 day apart for 1 week
clv.dyn <- fct.helper.create.clvdata.apparel.dyncov(
data.apparelTrans=rbind(apparelTrans,
data.table(Id=1, Price = 10,
Date = seq.Date(from=apparelTrans[Id==1, min(Date)]+1,
length.out = 8, by="1 day"))),
data.apparelDynCov=apparelDynCov,
estimation.split=NULL)
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv.dyn))
# every repeat-transaction is in walks (not lost)
# unique(tp.this.trans) because has a walk longer than 1 period (when tp.this.trans on week start)
expect_equal(l.walks$data.walks.trans.real[Id == 1, sort(unique(tp.this.trans))],
sort(clv.dyn@data.transactions[Id==1, Date][-1]))
# number of walks == num repeat transactions
# cannot check .N==8 because some walks may be longer than 1 period if second transaction on week start
expect_true(l.walks$data.walks.trans.real[Id == 1, uniqueN(walk_id)] == 8)
})
test_that("Real Trans Walk: None if there are no repeat transactions", {
skip_on_cran()
# Real trans walks only exist for repeat transactions
# if there are no repeat transactions, there should be no real trans walks
#
# Data: Remove repeat transactions of each customer from apparelTrans
# only keep first transaction of every customer
# Because in apparelTrans all customers have their transaction on the first date,
# no clvdata object can be created as it requires at least 1 period.
# Therefore move dates forward by up to 30days
apparelTrans.norepeat <- apparelTrans[order(Date), head(.SD, n=1), by="Id"]
apparelTrans.norepeat[, Date := Date + .I %% 30]
expect_silent(l.walks <- pnbd_dyncov_createwalks(fct.helper.create.clvdata.apparel.dyncov(apparelTrans.norepeat, apparelDynCov, estimation.split = NULL)))
expect_true(nrow(l.walks$data.walks.trans.real) == 0)
expect_true(ncol(l.walks$data.walks.trans.real) == 13)
expect_true(all(colnames(l.walks$data.walks.trans.real) == colnames(l.walks$data.walks.trans.aux)))
})
test_that("real life walk + aux life walk give original covariate data",{
skip_on_cran()
# use with estimation split because real+aux only exist in estimation period
clv.dyn <- fct.helper.create.clvdata.apparel.dyncov(data.apparelTrans=apparelTrans,
data.apparelDynCov=apparelDynCov,
estimation.split="2005-06-30")
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv.dyn))
fct.real.plus.aux <- function(id){
dt.plus <- rbind(l.walks$data.walks.life.real[Id == id],
l.walks$data.walks.life.aux[Id == id])
# no gap in cov dates (all 7 days apart)
expect_setequal(dt.plus[, as.numeric(diff(tp.cov.lower), units="days")], 7)
expect_setequal(dt.plus[, as.numeric(diff(tp.cov.upper), units="days")], 7)
# covs same as used to create
dt.original <- apparelDynCov[Id == id &
Cov.Date >= apparelTrans[Id == id, floor_date(min(Date), unit = "week")] &
Cov.Date <= lubridate::ymd("2005-06-30")]
expect_true(all(dt.plus[, .(Id, Cov.Date = tp.cov.lower, Marketing, Gender, Channel)] == dt.original))
}
# many are zerorep and do not have real trans walk
# fct.real.plus.aux("10")
# fct.real.plus.aux("100")
# fct.real.plus.aux("1000")
# all repeat buyers (zeroreps do not have real trans walk)
for(id in apparelTrans[, .N, by="Id"][N>1][1:25, Id]){
fct.real.plus.aux(id)
}
})
test_that("repeat buyers: life walk and first trans walk start on the same timepoint",{
skip_on_cran()
clv.dyn <- fct.helper.create.clvdata.apparel.dyncov(data.apparelTrans=apparelTrans,
data.apparelDynCov=apparelDynCov,
estimation.split="2005 06 30")
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv.dyn))
expect_true(l.walks$data.walks.life.real[Id == "10", min(tp.cov.lower)] == l.walks$data.walks.trans.real[Id == "10", min(tp.cov.lower)])
expect_true(l.walks$data.walks.life.real[Id == "100", min(tp.cov.lower)] == l.walks$data.walks.trans.real[Id == "100", min(tp.cov.lower)])
expect_true(l.walks$data.walks.life.real[Id == "1000", min(tp.cov.lower)] == l.walks$data.walks.trans.real[Id == "1000", min(tp.cov.lower)])
})
bachmannpatrick/CLVTools documentation built on Oct. 29, 2023, 2:16 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
skip_on_cran()
data("apparelTrans")
data("apparelDynCov")
test_that("interval created with eps same as if used shift()", {
skip_on_cran()
clv.time <- clv.time.weeks("ymd")
dt.interval <- pnbd_dyncov_covariate_add_interval_bounds(dt.cov=copy(apparelDynCov), clv.time=clv.time)
# same as if shifted + short manual
# no NAs in tp columns
dt.shift <- copy(apparelDynCov)
dt.shift[, tp.cov.lower := Cov.Date]
dt.shift[, tp.cov.upper := shift(tp.cov.lower, n=-1L), by="Id"]
dt.shift[, tp.cov.upper := tp.cov.upper - clv.time.epsilon(clv.time)]
dt.shift[is.na(tp.cov.upper), tp.cov.upper := tp.cov.lower + clv.time.number.timeunits.to.timeperiod(clv.time, user.number.periods=1L) - clv.time.epsilon(clv.time)]
expect_false(anyNA(dt.interval))
expect_false(anyNA(dt.shift))
expect_true(all(dt.interval == dt.shift))
})
test_that("Math for d_x is correct (Excel, tab Walk d)", {
skip_on_cran()
# From accompanying Excel, tab "Walk d"
clv.week <- clv.time.weeks("ymd")
# assuming Sunday (day 7) being the start of the week
oldval <- getOption("lubridate.week.start", 7)
options(lubridate.week.start=7)
# Monday = 2007-01-01
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-01")), 6/7)
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-02")), 5/7)
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-03")), 4/7)
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-04")), 3/7)
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-05")), 2/7)
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-06")), 1/7)
# Sunday, on boundary
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-07")), 1)
options(lubridate.week.start=oldval)
})
test_that("d_x changes correctly on (lower) boundary", {
skip_on_cran()
fct.expect.d.between.0.1 <- function(clv, d){
d <- clv.time.convert.user.input.to.timepoint(clv, user.timepoint = d)
expect_true(pnbd_dyncov_walk_d(clv.time=clv, tp.relevant.transaction = d) > 0)
expect_true(pnbd_dyncov_walk_d(clv.time=clv, tp.relevant.transaction = d) < 1)
}
fct.expect.d.equal.1 <- function(clv, d){
d <- clv.time.convert.user.input.to.timepoint(clv, user.timepoint = d)
expect_true(pnbd_dyncov_walk_d(clv.time=clv, tp.relevant.transaction = d) == 1)
}
# year
# 2006-12-31 < 1 and >0 (just before (lower) boundary)
# 2007-01-02: first after boundary, >0 and <1
# 2006-01-01: on (lower) boundary -> change -> 1
# 2007-01-01: on (lower) boundary -> 1
# 2008-01-01: on (lower) boundary -> 1 (also test for leap year)
clv.year <- clv.time.years("ymd")
fct.expect.d.between.0.1(clv.year, d="2005-12-31")
fct.expect.d.between.0.1(clv.year, d="2007-01-02")
fct.expect.d.equal.1(clv.year, d="2006-01-01")
fct.expect.d.equal.1(clv.year, d="2007-01-01")
fct.expect.d.equal.1(clv.year, d="2008-01-01")
# week
# 2007-01-01: monday, first day after (lower) boundary
# 2007-01-07: sunday, on (lower) boundary -> change -> 1
clv.week <- clv.time.weeks("ymd")
expect_equal(pnbd_dyncov_walk_d(clv.time=clv.week, tp.relevant.transaction = lubridate::ymd("2007-01-01")), 6/7)
fct.expect.d.equal.1(clv.week, d="2007-01-07")
# day
# ** TODO: changes always??, always == 1??
# clv.day <- clv.time.days("ymd")
# pnbd_dyncov_walk_d(clv.time=clv.day, tp.relevant.transaction = lubridate::ymd("2007-01-01"))
# hour
# 00:00:01: < 1 and > 0 (first after boundary)
# 00:00:59: < 1
# 00:01:00: < 1
# 00:59:00: < 1
# 00:59:59: < 1 and > 0 (just before boundary)
# 00:00:00: == 1
# 00:01:00: == 1
clv.hour <- clv.time.hours("ymd HMS")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:00:01")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:00:59")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:01:00")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:59:00")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:59:59")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 00:59:59")
fct.expect.d.between.0.1(clv.hour, d="2007-01-01 23:59:59")
# on lower boundary
fct.expect.d.equal.1(clv.hour, d="2007-01-01 00:00:00")
fct.expect.d.equal.1(clv.hour, d="2007-01-01 01:00:00")
fct.expect.d.equal.1(clv.hour, d="2007-01-01 23:00:00")
fct.expect.d.equal.1(clv.hour, d="2007-12-31 23:00:00")
})
test_that("d_omega is actually correct in cbs", {
skip_on_cran()
apparelTrans.spaced <- copy(apparelTrans)
# Fix data: space customers' first transaction out over whole first week
ids <- apparelTrans.spaced[, unique(Id)]
apparelTrans.spaced[, first.trans := Date == min(Date), by="Id"]
tp.first.trans <- apparelTrans.spaced[, min(Date)]
for(i in 1:7){
apparelTrans.spaced[Id == ids[i] & first.trans == TRUE, Date := tp.first.trans+days(i-1)]
}
clv.dyn.spaced <- fct.helper.create.clvdata.apparel.dyncov(apparelTrans.spaced, apparelDynCov, estimation.split = 38)
dt.cbs <- pnbd_dyncov_cbs(clv.dyn.spaced)
# We assume Sunday (7) to be start
dt.cbs[, weekday := lubridate::wday(date.first.actual.trans, label=TRUE, abbr=TRUE)]
dt.cbs[weekday == "Mon", unique(d_omega) == 6/7]
dt.cbs[weekday == "Tue", unique(d_omega) == 5/7]
dt.cbs[weekday == "Wed", unique(d_omega) == 4/7]
dt.cbs[weekday == "Thu", unique(d_omega) == 3/7]
dt.cbs[weekday == "Fri", unique(d_omega) == 2/7]
dt.cbs[weekday == "Sat", unique(d_omega) == 1/7]
dt.cbs[weekday == "Sun", unique(d_omega) == 1]
})
test_that("tjk is correctly calculated", {
skip_on_cran()
# Fake transactions for customer 1
# - dates: 2005-01-03, 2005-01-04, 2005-01-05, 2005-01-07, 2005-01-18
# - Id: Replace existing transactions of "1" because also requires covariates
#
# 2005-01-03 -> 2005-01-04: tjk=1/7
# 2005-01-04 -> 2005-01-05: tjk=1/7
# 2005-01-05 -> 2005-01-07: tjk=2/7
# 2005-01-07 -> 2005-01-18: tjk=11/7
apparelTrans.tjk <- rbindlist(list(apparelTrans[Id != "1"],
data.table(Id="1", Date=lubridate::ymd(c("2005-01-03", "2005-01-04", "2005-01-05", "2005-01-07", "2005-01-18")), Price=12.34)))
expect_silent(l.walks <- pnbd_dyncov_createwalks(fct.helper.create.clvdata.apparel.dyncov(apparelTrans.tjk, apparelDynCov, estimation.split = NULL)))
expect_equal(unique(l.walks$data.walks.trans.real[Id == "1", c("tp.this.trans", "tjk")])[order(tp.this.trans)],
data.table(tp.this.trans=lubridate::ymd(c("2005-01-04", "2005-01-05", "2005-01-07", "2005-01-18")),
tjk=c(1/7, 1/7, 2/7, 11/7)))
})
test_that("tjk is correctly calculated for aux trans with t.x=Tcal", {
skip_on_cran()
tp.T <- lubridate::ymd("2006-01-01")
# ensure at least one with t.x=T
apparelTrans.zeroaux <- rbindlist(list(apparelTrans, data.table(Id="1", Date=tp.T, Price=12.34)))
clv.dyn.zeroaux <- fct.helper.create.clvdata.apparel.dyncov(apparelTrans.zeroaux, apparelDynCov, estimation.split = tp.T)
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv.dyn.zeroaux))
ids.tx.eq.T <- apparelTrans.zeroaux[Date == tp.T, Id]
expect_true(l.walks$data.walks.trans.aux[Id %in% ids.tx.eq.T, all(tjk == 0)])
# Coincidence: T is also on the boundary!
expect_true(l.walks$data.walks.trans.aux[Id %in% ids.tx.eq.T, all(d1 == 1)])
})
test_that("Aux walk splitting method correct", {
# this is not from test case excel
skip_on_cran()
clv.time <- clv.time.weeks("ymd")
# Start of week: Monday = 2007-01-01
dt.cov <- data.table(Id="1",
tp.cov.lower = lubridate::ymd(c("2007-01-01", "2007-01-08", "2007-01-15")),
tp.cov.upper = lubridate::ymd(c("2007-01-07", "2007-01-14", "2007-01-21")),
cov.data = c(0.123, 0.678, 2.345))
fct.get.walk <- function(tp.last.trans, tp.estimation.end){
dt.tp.first.last <- data.table(tp.last.trans = lubridate::ymd(tp.last.trans), Id="1")
clv.time@timepoint.estimation.end <- lubridate::ymd(tp.estimation.end)
return(pnbd_dyncov_createwalks_auxwalk(dt.cov, dt.tp.first.last, names.cov="cov.data", clv.time = clv.time))
}
dt.walk <- fct.get.walk(tp.last.trans = "2007-01-05", tp.estimation.end = "2007-01-07")
expect_true(dt.walk[, .N] == 1)
expect_true(dt.walk[, cov.data] == 0.123)
dt.walk <- fct.get.walk(tp.last.trans = "2007-01-05", tp.estimation.end = "2007-01-08")
expect_true(dt.walk[, .N] == 2)
expect_true(all(dt.walk[, cov.data] == c(0.123, 0.678)))
})
test_that("Aux walk is 2 periods if T is on week start and alive at T-1 one day before (and has no real life walk)", {
# T on week start, alive is T-1 -> aux walk is 2 periods, but no real life walk (run test with T on all 7 days)
for(i in seq(7)){
data.trans <- copy(apparelTrans)
data.cov <- copy(apparelDynCov)
oldval <- getOption("lubridate.week.start")
options(lubridate.week.start=(7+i-1) %% 7+1)
# print(getOption("lubridate.week.start"))
# make cov dates to be on week start
data.cov[, Cov.Date := Cov.Date + i]
# split is on start of week
date.estimation.split <- lubridate::ymd("2005-06-26") + i
# Id==1 is zero-repeater
data.trans[Id==1, Date := date.estimation.split-1]
clv <- fct.helper.create.clvdata.apparel.dyncov(data.apparelTrans=data.trans,
data.apparelDynCov=data.cov,
estimation.split=date.estimation.split)
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv))
# has 2 aux walks
expect_true(l.walks$data.walks.life.aux[Id==1, .N] == 2)
# no real walk
expect_true(l.walks$data.walks.life.real[Id==1, .N] == 0)
options(lubridate.week.start=oldval)
}
})
test_that("Aux walks not lost if there are covariates only for the calibration period (see #134)", {
skip_on_cran()
clv.short <- fct.helper.create.clvdata.apparel.dyncov(data.apparelTrans=apparelTrans[Date <= "2005-12-31"],
data.apparelDynCov=apparelDynCov[Cov.Date <= "2005-12-31"],
estimation.split=NULL)
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv.short))
expect_true(l.walks$data.walks.life.aux[, uniqueN(Id)] == 250)
expect_true(l.walks$data.walks.trans.aux[, uniqueN(Id)] == 250)
})
test_that("Real Trans Walk correct", {
skip_on_cran()
# Fake transactions for customer "1"
# - Id: Replace existing transactions of "1" because also requires covariates
# - on and off the cov boundary
# - cov boundaries: [2005-01-02, 2005-01-08][2005-01-09, 2005-01-15][2005-01-16, 2005-01-22][2005-01-23, 2005-01-29][2005-01-30, 2005-02-05]
# - cov values: 0.123 0.234 0.345 0.456 0.567
# - transaction dates: 2005-01-02, 2005-01-07, 2005-01-10, 2005-01-15, 2005-01-16, 2005-01-30
#
# Real Trans Walk: Covariates up to a repeat transaction
# Resulting real trans walks
# 2005-01-07: [0.123]
# 2005-01-10: [0.123], [0.234]
# 2005-01-15: [0.234]
# 2005-01-16: [0.234], [0.345]
# 2005-01-30: [0.345], [0.456], [0.567]
#
apparelTrans.realwalk <- rbindlist(list(apparelTrans[Id != "1"],
data.table(Id="1", Date=lubridate::ymd(c("2005-01-02", "2005-01-07", "2005-01-10", "2005-01-15", "2005-01-16", "2005-01-30")), Price=12.34)))
dates.cov.realwalk <- lubridate::ymd(c("2005-01-02", "2005-01-09", "2005-01-16", "2005-01-23", "2005-01-30"))
values.cov.realwalk <- c(0.123, 0.234, 0.345, 0.456, 0.567)
cov.realwalk <- rbindlist(list(apparelDynCov[!(Id == "1" & Cov.Date %in% dates.cov.realwalk)],
data.table(Id="1", Cov.Date=dates.cov.realwalk, Marketing=c(0.123, 0.234, 0.345, 0.456, 0.567), Gender=0, Channel=0)))
expect_silent(l.walks <- pnbd_dyncov_createwalks(fct.helper.create.clvdata.apparel.dyncov(apparelTrans.realwalk, cov.realwalk, estimation.split = NULL)))
expect_equal(
l.walks$data.walks.trans.real[Id==1][order(tp.this.trans), list(tp.this.trans=as.character(tp.this.trans), Marketing)],
rbindlist(list(
data.table(tp.this.trans="2005-01-07", Marketing=0.123),
data.table(tp.this.trans="2005-01-10", Marketing=c(0.123, 0.234)),
data.table(tp.this.trans="2005-01-15", Marketing=0.234),
data.table(tp.this.trans="2005-01-16", Marketing=c(0.234, 0.345)),
data.table(tp.this.trans="2005-01-30", Marketing=c(0.345, 0.456, 0.567)))))
})
test_that("All walks have basic correctness, estimation.split at every day of week", {
skip_on_cran()
# From walk creation asserts function
# Test with estimation.split at every week day
for(i in seq(7)){
clv <- fct.helper.create.clvdata.apparel.dyncov(data.apparelTrans=apparelTrans,
data.apparelDynCov=apparelDynCov,
# move split by i days
estimation.split=lubridate::ymd("2005-06-30") + i)
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv))
dt.cbs <- pnbd_dyncov_cbs(clv)
fct.walk.basic.correctness <- function(dt.walks){
# no NA# coreelements
expect_true(setequal(key(dt.walks), c("Id", "walk_id", "tp.this.trans", "tp.cov.lower")))
expect_true(dt.walks[, !is.unsorted(abs_pos, strictly = TRUE)])
expect_true(dt.walks[, !is.unsorted(walk_id, strictly = FALSE)])
expect_true(dt.walks[, !is.unsorted(walk_from, strictly = FALSE)])
expect_true(dt.walks[, !is.unsorted(walk_to, strictly = FALSE)])
expect_true(!anyNA(dt.walks))
# walks are backwards looking from second transaction: transaction must be after all of walks' lower covariate periods
# all transactions must be after all of walks' lower covariate periods
}
fct.walk.basic.correctness(l.walks$data.walks.life.aux)
fct.walk.basic.correctness(l.walks$data.walks.life.real)
fct.walk.basic.correctness(l.walks$data.walks.trans.aux)
fct.walk.basic.correctness(l.walks$data.walks.trans.real)
expect_true(l.walks$data.walks.trans.real[tp.this.trans < tp.cov.lower, .N] == 0)
# All transactions on week start are longer than 1 walk (because also include cov of period before)
l.walks$data.walks.trans.real[wday(tp.this.trans) == getOption("lubridate.week.start", 1), walk_id]
# number of real trans walks == num repeat transactions
# all aux walks:
# every Id only once
# every Id in walks
# same length per customer
expect_true(l.walks$data.walks.life.aux[, list(num_walks = uniqueN(walk_id)), by="Id"][, all(num_walks==1)])
expect_true(l.walks$data.walks.life.aux[, uniqueN(Id)] == apparelTrans[, uniqueN(Id)])
expect_true(setequal(l.walks$data.walks.life.aux[, unique(Id)], apparelTrans[, unique(Id)]))
expect_true(l.walks$data.walks.trans.aux[, list(num_walks = uniqueN(walk_id)), by="Id"][, all(num_walks==1)])
expect_true(l.walks$data.walks.trans.aux[, uniqueN(Id)] == apparelTrans[, uniqueN(Id)])
expect_true(setequal(l.walks$data.walks.trans.aux[, unique(Id)], apparelTrans[, unique(Id)]))
expect_true(identical(l.walks$data.walks.life.aux[, .N, keyby="Id"],
l.walks$data.walks.trans.aux[, .N, keyby="Id"]))
# lifetime aux walk:
# no date overlap with real lifetime walks
dt.tmp <- l.walks$data.walks.life.aux[, list(first_cov_aux = min(tp.cov.lower)), keyby="Id"]
dt.tmp[l.walks$data.walks.life.real[, list(first_cov_real = min(tp.cov.lower)) , keyby="Id"], first_cov_real := i.first_cov_real, on="Id"]
# some first_cov_real are NA because have no real walk
expect_true(dt.tmp[first_cov_aux < first_cov_real, .N] == 0)
# lifetime real walk:
# exactly 1 walk per customer
# all ids except where aux walk reaches to the first transactions (coming alive)
# (number of customers = num customers in trans real walks where .N>1)
# n real walk + n aux walk >= ceiling(Tcal)
expect_true(l.walks$data.walks.life.real[, list(num_walks=uniqueN(walk_id)), keyby="Id"][, all(num_walks == 1)])
# dt.tmp[clv.fitted@clv.data@data.transactions[, list(last_trans = max(Date)), by="Id"], last_trans := i.last_trans, on="Id"]
dt.tmp[dt.cbs, first_trans := i.date.first.actual.trans, on="Id"]
expect_true(setequal(l.walks$data.walks.life.real[, unique(Id)],
dt.tmp[first_cov_aux > first_trans, Id]))
# trans real walks:
# every Id with x>0 is in ...
# ... with x-1 walks
expect_true(setequal(l.walks$data.walks.trans.real[, unique(Id)], dt.cbs[x>0, Id]))
expect_true(identical(l.walks$data.walks.trans.real[, list(num_walks=as.double(uniqueN(walk_id))), keyby="Id"],
dt.cbs[x>0, list(num_walks=as.double(x)), keyby="Id"]))
# trans walks
# tjk >= 0 (== 0 when t.x=T)
expect_true(l.walks$data.walks.trans.aux[tjk < 0, .N] == 0)
expect_true(l.walks$data.walks.trans.real[tjk < 0, .N] == 0)
# d1 and d_omega measures are in (0,1]
expect_true(dt.cbs[, all(d_omega > 0 & d_omega <= 1)])
expect_true(l.walks$data.walks.trans.aux[, all(d1 > 0 & d1 <= 1)])
expect_true(l.walks$data.walks.trans.real[, all(d1 > 0 & d1 <= 1)])
}
})
test_that("Real Trans Walk do no lose walk if transactions only 1 eps apart", {
skip_on_cran()
# real trans walk gives walk and not lost if transactions only 1 eps apart
# Customer 1 is zero repeater. Add 8 transactions only 1 day apart for 1 week
clv.dyn <- fct.helper.create.clvdata.apparel.dyncov(
data.apparelTrans=rbind(apparelTrans,
data.table(Id=1, Price = 10,
Date = seq.Date(from=apparelTrans[Id==1, min(Date)]+1,
length.out = 8, by="1 day"))),
data.apparelDynCov=apparelDynCov,
estimation.split=NULL)
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv.dyn))
# every repeat-transaction is in walks (not lost)
# unique(tp.this.trans) because has a walk longer than 1 period (when tp.this.trans on week start)
expect_equal(l.walks$data.walks.trans.real[Id == 1, sort(unique(tp.this.trans))],
sort(clv.dyn@data.transactions[Id==1, Date][-1]))
# number of walks == num repeat transactions
# cannot check .N==8 because some walks may be longer than 1 period if second transaction on week start
expect_true(l.walks$data.walks.trans.real[Id == 1, uniqueN(walk_id)] == 8)
})
test_that("Real Trans Walk: None if there are no repeat transactions", {
skip_on_cran()
# Real trans walks only exist for repeat transactions
# if there are no repeat transactions, there should be no real trans walks
#
# Data: Remove repeat transactions of each customer from apparelTrans
# only keep first transaction of every customer
# Because in apparelTrans all customers have their transaction on the first date,
# no clvdata object can be created as it requires at least 1 period.
# Therefore move dates forward by up to 30days
apparelTrans.norepeat <- apparelTrans[order(Date), head(.SD, n=1), by="Id"]
apparelTrans.norepeat[, Date := Date + .I %% 30]
expect_silent(l.walks <- pnbd_dyncov_createwalks(fct.helper.create.clvdata.apparel.dyncov(apparelTrans.norepeat, apparelDynCov, estimation.split = NULL)))
expect_true(nrow(l.walks$data.walks.trans.real) == 0)
expect_true(ncol(l.walks$data.walks.trans.real) == 13)
expect_true(all(colnames(l.walks$data.walks.trans.real) == colnames(l.walks$data.walks.trans.aux)))
})
test_that("real life walk + aux life walk give original covariate data",{
skip_on_cran()
# use with estimation split because real+aux only exist in estimation period
clv.dyn <- fct.helper.create.clvdata.apparel.dyncov(data.apparelTrans=apparelTrans,
data.apparelDynCov=apparelDynCov,
estimation.split="2005-06-30")
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv.dyn))
fct.real.plus.aux <- function(id){
dt.plus <- rbind(l.walks$data.walks.life.real[Id == id],
l.walks$data.walks.life.aux[Id == id])
# no gap in cov dates (all 7 days apart)
expect_setequal(dt.plus[, as.numeric(diff(tp.cov.lower), units="days")], 7)
expect_setequal(dt.plus[, as.numeric(diff(tp.cov.upper), units="days")], 7)
# covs same as used to create
dt.original <- apparelDynCov[Id == id &
Cov.Date >= apparelTrans[Id == id, floor_date(min(Date), unit = "week")] &
Cov.Date <= lubridate::ymd("2005-06-30")]
expect_true(all(dt.plus[, .(Id, Cov.Date = tp.cov.lower, Marketing, Gender, Channel)] == dt.original))
}
# many are zerorep and do not have real trans walk
# fct.real.plus.aux("10")
# fct.real.plus.aux("100")
# fct.real.plus.aux("1000")
# all repeat buyers (zeroreps do not have real trans walk)
for(id in apparelTrans[, .N, by="Id"][N>1][1:25, Id]){
fct.real.plus.aux(id)
}
})
test_that("repeat buyers: life walk and first trans walk start on the same timepoint",{
skip_on_cran()
clv.dyn <- fct.helper.create.clvdata.apparel.dyncov(data.apparelTrans=apparelTrans,
data.apparelDynCov=apparelDynCov,
estimation.split="2005 06 30")
expect_silent(l.walks <- pnbd_dyncov_createwalks(clv.dyn))
expect_true(l.walks$data.walks.life.real[Id == "10", min(tp.cov.lower)] == l.walks$data.walks.trans.real[Id == "10", min(tp.cov.lower)])
expect_true(l.walks$data.walks.life.real[Id == "100", min(tp.cov.lower)] == l.walks$data.walks.trans.real[Id == "100", min(tp.cov.lower)])
expect_true(l.walks$data.walks.life.real[Id == "1000", min(tp.cov.lower)] == l.walks$data.walks.trans.real[Id == "1000", min(tp.cov.lower)])
})
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.