In hnguyen1174/online_visits_assessment: To assess the effectiveness of online visit on booking metrics

In this notebook, I performed statistical tests as if the dataset was generated from a randomized controlled experiment (such as an AB Test).

1. Load Libraries

library(tidyverse)
library(here)
library(lubridate)
library(janitor)
library(zoo)
library(ggthemes)

library(forecast)
library(stargazer)
library(data.table)
library(tableone)
library(lattice)
library(pwr)
library(rcompanion)
library(scales)
library(plm)
library(readxl)
library(MatchIt)
library(lfe)
library(Synth)
library(gsynth)
library(panelView)
library(CausalImpact)
library(optmatch)

library(zeallot)

devtools::load_all()

2. Load and Process Data

c(metrics, property) %<-% get_property_data()

3. T-test (aggregated)

control_ids <- property %>% 
  filter(treat == 0) %>% 
  pull(property_id) %>% 
  unique()

df_for_test <- metrics %>% 
  mutate(treatment = if_else(property_id %in% control_ids, 0, 1)) %>% 
  mutate(post = if_else(date < as.Date('2019-01-01'), 'Pre', 'Post')) %>% 
  group_by(property_id, post, treatment) %>% 
  summarize(avg_bookings = mean(bookings),
            avg_booking_amount = mean(booking_amount),
            avg_booking_requests = mean(booking_requests),
            avg_property_page_views_per_visit = mean(property_page_views_visits),
            avg_displays_per_searchpage = mean(number_of_times_the_property_got_displayed_in_search_page)) %>% 
  ungroup() %>% 
  filter(post == 'Post')

# 690
control <- df_for_test %>% 
  filter(treatment == 0)

# 1294
treatment <- df_for_test %>% 
  filter(treatment == 1) %>% 
  sample_n(nrow(control))

df_for_test_resample <- control %>% 
  bind_rows(treatment)

Results: avg_bookings: not stat sig avg_booking_amount: not stat sig avg_booking_requests: stat sig (-0.00500467) avg_property_page_views_per_visit: stat sig (-1.85892) * avg_displays_per_searchpage: not stat sig

# avg_bookings
# avg_booking_amount
# avg_booking_requests
# avg_property_page_views_per_visit
# avg_displays_per_searchpage

t.test(avg_displays_per_searchpage ~ treatment, 
       data = df_for_test_resample,
       var.equal = FALSE, 
       alternative = 'greater')

3. T-test (by location)

3.1. Texas

df_for_test <- metrics %>% 
  left_join(dplyr::select(property, property_id, location)) %>% 
  mutate(treatment = if_else(property_id %in% control_ids, 0, 1)) %>% 
  mutate(post = if_else(date < as.Date('2019-01-01'), 'Pre', 'Post')) %>% 
  group_by(property_id, location, post, treatment) %>% 
  summarize(avg_bookings = mean(bookings),
            avg_booking_amount = mean(booking_amount),
            avg_booking_requests = mean(booking_requests),
            avg_property_page_views_per_visit = mean(property_page_views_visits),
            avg_displays_per_searchpage = mean(number_of_times_the_property_got_displayed_in_search_page)) %>% 
  ungroup() %>% 
  filter(post == 'Post') %>% 
  filter(location == 'Austin, Texas')

control <- df_for_test %>% 
  filter(treatment == 0)

treatment <- df_for_test %>% 
  filter(treatment == 1) %>% 
  sample_n(nrow(control))

df_for_test_resample <- control %>% 
  bind_rows(treatment)

Results for Texas: avg_bookings: stat sig (-0.00898518) avg_booking_amount: not stat sig avg_booking_requests: stat sig (-0.01073499) avg_property_page_views_per_visit: not stat sig * avg_displays_per_searchpage: not stat sig

# avg_bookings
# avg_booking_amount
# avg_booking_requests
# avg_property_page_views_per_visit
# avg_displays_per_searchpage

t.test(avg_displays_per_searchpage ~ treatment, 
       data = df_for_test_resample,
       var.equal = FALSE, 
       alternative = 'greater')

3.2. Nevada

df_for_test <- metrics %>% 
  left_join(dplyr::select(property, property_id, location)) %>% 
  mutate(treatment = if_else(property_id %in% control_ids, 0, 1)) %>% 
  mutate(post = if_else(date < as.Date('2019-01-01'), 'Pre', 'Post')) %>% 
  group_by(property_id, location, post, treatment) %>% 
  summarize(avg_bookings = mean(bookings),
            avg_booking_amount = mean(booking_amount),
            avg_booking_requests = mean(booking_requests),
            avg_property_page_views_per_visit = mean(property_page_views_visits),
            avg_displays_per_searchpage = mean(number_of_times_the_property_got_displayed_in_search_page)) %>% 
  ungroup() %>% 
  filter(post == 'Post') %>% 
  filter(location == 'Vegas, Nevada')

control <- df_for_test %>% 
  filter(treatment == 0)

treatment <- df_for_test %>% 
  filter(treatment == 1) %>% 
  sample_n(nrow(control))

df_for_test_resample <- control %>% 
  bind_rows(treatment)

Results for Vegas: avg_bookings: not stat sig avg_booking_amount: weak stat sig (-25.36331) avg_booking_requests: weak stat sig (-0.00977751) avg_property_page_views_per_visit: not stat sig * avg_displays_per_searchpage: not stat sig

# avg_bookings
# avg_booking_amount
# avg_booking_requests
# avg_property_page_views_per_visit
# avg_displays_per_searchpage

t.test(avg_displays_per_searchpage ~ treatment, 
       data = df_for_test_resample,
       var.equal = FALSE, 
       alternative = 'greater')

3.3. Florida

df_for_test <- metrics %>% 
  left_join(dplyr::select(property, property_id, location)) %>% 
  mutate(treatment = if_else(property_id %in% control_ids, 0, 1)) %>% 
  mutate(post = if_else(date < as.Date('2019-01-01'), 'Pre', 'Post')) %>% 
  group_by(property_id, location, post, treatment) %>% 
  summarize(avg_bookings = mean(bookings),
            avg_booking_amount = mean(booking_amount),
            avg_booking_requests = mean(booking_requests),
            avg_property_page_views_per_visit = mean(property_page_views_visits),
            avg_displays_per_searchpage = mean(number_of_times_the_property_got_displayed_in_search_page)) %>% 
  ungroup() %>% 
  filter(post == 'Post') %>% 
  filter(location == 'Destin, Florida')

control <- df_for_test %>% 
  filter(treatment == 0)

treatment <- df_for_test %>% 
  filter(treatment == 1) %>% 
  sample_n(nrow(control))

df_for_test_resample <- control %>% 
  bind_rows(treatment)

Results for Florida: avg_bookings: not stat sig avg_booking_amount: not stat sig avg_booking_requests: not stat sig avg_property_page_views_per_visit: not stat sig * avg_displays_per_searchpage: not stat sig

# avg_bookings
# avg_booking_amount
# avg_booking_requests
# avg_property_page_views_per_visit
# avg_displays_per_searchpage

t.test(avg_displays_per_searchpage ~ treatment, 
       data = df_for_test_resample,
       var.equal = FALSE, 
       alternative = 'greater')

hnguyen1174/online_visits_assessment documentation built on Dec. 20, 2021, 4:46 p.m.