rm(list = ls())
library(tidyverse)

## ── Attaching packages ─────────────────────────────────────── tidyverse 1.3.1 ──

## ✓ ggplot2 3.3.5     ✓ purrr   0.3.4
## ✓ tibble  3.1.4     ✓ dplyr   1.0.7
## ✓ tidyr   1.1.3     ✓ stringr 1.4.0
## ✓ readr   2.0.1     ✓ forcats 0.5.1

## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## x dplyr::filter() masks stats::filter()
## x dplyr::lag()    masks stats::lag()

library(data.table)

## 
## Attaching package: 'data.table'

## The following objects are masked from 'package:dplyr':
## 
##     between, first, last

## The following object is masked from 'package:purrr':
## 
##     transpose

library(patchwork)
library(glue)

## 
## Attaching package: 'glue'

## The following object is masked from 'package:dplyr':
## 
##     collapse

library(correlation)

## Warning: package 'correlation' was built under R version 4.1.2

library(GGally)

## Registered S3 method overwritten by 'GGally':
##   method from   
##   +.gg   ggplot2

theme_set(theme_minimal())

d0 <- fread("../data_ukraine/user_activity.csv")
d0 <- mutate_if(d0, is.integer, as.numeric)
setDT(d0)
d0[, lineitem := paste0("adgroup", group-1)]
glimpse(d0)

## Rows: 17,049
## Columns: 16
## $ id_str           <int64> 1470103877282320391, 1424873675795308555, 92507535,…
## $ subcamp01_tweets <dbl> 58, 19, 0, 22, 0, 66, 0, 0, 30, 2, 0, 15, 30, 20, 0, …
## $ subcamp02_tweets <dbl> 86, 0, 0, 7, 0, 112, 0, 0, 25, 2, 0, 131, 3, 17, 2, 2…
## $ subcamp03_tweets <dbl> 54, 0, 0, 0, 0, 201, 0, 5, 30, 1, 5, 71, 9, 31, 1, 8,…
## $ subcamp04_tweets <dbl> 79, 0, 0, 14, 0, 149, 0, 7, 18, 4, 0, 79, 10, 41, 3, …
## $ subcamp05_tweets <dbl> 40, 0, 7, 10, 0, 68, 0, 0, 31, 3, 0, 76, 5, 30, 2, 8,…
## $ subcamp06_tweets <dbl> 20, 0, 1, 3, 0, 108, 0, 1, 14, 0, 0, 16, 5, 21, 0, 7,…
## $ subcamp07_tweets <dbl> 29, 0, 0, 2, 0, 57, 0, 0, 19, 0, 0, 1, 1, 9, 0, 3, 23…
## $ subcamp08_tweets <dbl> 44, 0, 6, 1, 0, 91, 0, 1, 2, 0, 0, 14, 16, 13, 1, 12,…
## $ subcamp09_tweets <dbl> 46, 0, 1, 0, 0, 63, 0, 1, 0, 1, 0, 34, 6, 16, 1, 0, 1…
## $ subcamp10_tweets <dbl> 72, 0, 4, 2, 0, 57, 0, 0, 15, 2, 3, 10, 3, 15, 2, 8, …
## $ subcamp11_tweets <dbl> 53, 0, 6, 6, 0, 125, 0, 0, 30, 4, 1, 28, 3, 10, 2, 6,…
## $ group            <dbl> 8, 8, 12, 11, 12, 24, 8, 8, 8, 8, 10, 8, 8, 8, 12, 7,…
## $ condition        <chr> "t", "t", "c", "t", "t", "t", "t", "c", "c", "c", "c"…
## $ block            <chr> "08_108", "08_128", "12_55", "11_25", "12_26", "24_82…
## $ lineitem         <chr> "adgroup7", "adgroup7", "adgroup11", "adgroup10", "ad…

d0[, .(group, lineitem)][order(group)] |> distinct()

##     group  lineitem
##  1:     1  adgroup0
##  2:     2  adgroup1
##  3:     3  adgroup2
##  4:     4  adgroup3
##  5:     5  adgroup4
##  6:     6  adgroup5
##  7:     7  adgroup6
##  8:     8  adgroup7
##  9:     9  adgroup8
## 10:    10  adgroup9
## 11:    11 adgroup10
## 12:    12 adgroup11
## 13:    13 adgroup12
## 14:    14 adgroup13
## 15:    15 adgroup14
## 16:    16 adgroup15
## 17:    17 adgroup16
## 18:    18 adgroup17
## 19:    19 adgroup18
## 20:    20 adgroup19
## 21:    21 adgroup20
## 22:    22 adgroup21
## 23:    23 adgroup22
## 24:    24 adgroup23
## 25:    25 adgroup24
## 26:    26 adgroup25
## 27:    27 adgroup26
##     group  lineitem

d0[, .N, keyby = .(condition)]

##    condition    N
## 1:         c 8561
## 2:         t 8488

d_groupn <- d0[, .N, keyby = .(group, lineitem, condition)]
d_groupn

##     group  lineitem condition   N
##  1:     1  adgroup0         c 290
##  2:     1  adgroup0         t 292
##  3:     2  adgroup1         c 363
##  4:     2  adgroup1         t 356
##  5:     3  adgroup2         c 387
##  6:     3  adgroup2         t 383
##  7:     4  adgroup3         c 371
##  8:     4  adgroup3         t 363
##  9:     5  adgroup4         c 227
## 10:     5  adgroup4         t 224
## 11:     6  adgroup5         c 305
## 12:     6  adgroup5         t 301
## 13:     7  adgroup6         c 251
## 14:     7  adgroup6         t 243
## 15:     8  adgroup7         c 412
## 16:     8  adgroup7         t 409
## 17:     9  adgroup8         c 377
## 18:     9  adgroup8         t 370
## 19:    10  adgroup9         c 381
## 20:    10  adgroup9         t 375
## 21:    11 adgroup10         c 227
## 22:    11 adgroup10         t 229
## 23:    12 adgroup11         c 274
## 24:    12 adgroup11         t 269
## 25:    13 adgroup12         c 361
## 26:    13 adgroup12         t 360
## 27:    14 adgroup13         c 348
## 28:    14 adgroup13         t 349
## 29:    15 adgroup14         c 271
## 30:    15 adgroup14         t 275
## 31:    16 adgroup15         c 398
## 32:    16 adgroup15         t 401
## 33:    17 adgroup16         c 305
## 34:    17 adgroup16         t 300
## 35:    18 adgroup17         c 336
## 36:    18 adgroup17         t 338
## 37:    19 adgroup18         c 386
## 38:    19 adgroup18         t 387
## 39:    20 adgroup19         c 400
## 40:    20 adgroup19         t 394
## 41:    21 adgroup20         c 220
## 42:    21 adgroup20         t 227
## 43:    22 adgroup21         c 386
## 44:    22 adgroup21         t 366
## 45:    23 adgroup22         c 225
## 46:    23 adgroup22         t 232
## 47:    24 adgroup23         c 352
## 48:    24 adgroup23         t 345
## 49:    25 adgroup24         c 290
## 50:    25 adgroup24         t 294
## 51:    26 adgroup25         c 204
## 52:    26 adgroup25         t 199
## 53:    27 adgroup26         c 214
## 54:    27 adgroup26         t 207
##     group  lineitem condition   N

d1 <- melt(d0, id.vars = c("id_str", "group", "lineitem", "condition", "block"), variable.name = "campaign", value.name = "active")

d2 <- d1 %>% filter(active > 0) %>% group_by(group, lineitem, condition, campaign) %>% summarise(active_user_count = n(), 
                                                                                                 avg_active_tweets = mean(active), 
                                                                                                 super_user_count = sum(active >= 500), 
                                                                                                 idle_user_count = sum(active <= 2))

## `summarise()` has grouped output by 'group', 'lineitem', 'condition'. You can override using the `.groups` argument.

# d2 <- d1[active > 0, .(active_user_count = .N), keyby = .(group, lineitem, condition, campaign)]
d3 <- left_join(d2, d_groupn)

## Joining, by = c("group", "lineitem", "condition")

setDT(d3)
d3[, active_prop := active_user_count / N]
d3[, super_user_prop := super_user_count / N]
d3[, idle_user_prop := idle_user_count / N]

d1[d1$active > 500, ][order(-d1$active), ] %>%
    ggplot(aes(active)) +
    geom_histogram(bins = 100) +
    labs(x = "tweets per day (only > 500 tweets per day)", y = "no. of users")

## Warning: Removed 183948 rows containing non-finite values (stat_bin).

ggplot(d1, aes(active, col = campaign)) +
    stat_ecdf(size = 1) +
    labs(x = "no. of tweets", y = "proportion of users")

reach <- list.files("../campaign-adgroup/ukraine/export_data", pattern = "*-total.csv", full.names = T)
reach <- reach[!str_detect(reach, "06-28|06-30|07-01")]

campaigns <- d3[, unique(campaign)]

change_ad_group_name <- function(x){
  if (substring(x, 8, 8) == "0"){
    return (paste0(substring(x, 1, 7), substring(x, 9, 9)))
  }
  return (x)
}

plots <- list()
c <- 1
for (c in 1:length(campaigns)) {
    camp <- campaigns[c]
    tempdat <- d3[campaign == camp & condition == "t"]
    temp_reach <- fread(reach[c])
    if (c >= 6){
      temp_reach <- temp_reach %>% filter(Objective == "Video views") %>% filter(`Ad Group status` == "Running")
      temp_reach$`Ad Group name` <- sapply(temp_reach$`Ad Group name`, function(x) change_ad_group_name(x))
    }
    temp_reach <- temp_reach[, .(`Ad Group name`, `Total audience reach`)]
    setnames(temp_reach, c("lineitem", "reach"))
    tempdat <- left_join(tempdat, temp_reach)
    setDT(tempdat)
    tempdat[, reach := as.numeric(reach)]
    tempdat$reach[is.na(tempdat$reach)] <- 0
    tempdat[, reach_prop := reach / N]
    ord <- tempdat[order(reach_prop), lineitem]
    tempdat[, lineitem := factor(lineitem, levels = ord)]
    
    p1 <- ggplot(tempdat, aes(active_prop, lineitem)) +
        geom_bar(stat = "identity", alpha = 0.3) +
        geom_point(aes(x = reach_prop)) +
        geom_text(aes(label = paste0(active_user_count, "/", N)), size = 3) +
        scale_x_continuous(limits = c(0, 1)) +
        labs(x = "proportion active users (treatment)", y = "", title = glue("day {c}"))
    p1
    plots[[c]] <- p1
}

## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"

plots

## [[1]]

## 
## [[2]]

## 
## [[3]]

## 
## [[4]]

## 
## [[5]]

## 
## [[6]]

## 
## [[7]]

## 
## [[8]]

## 
## [[9]]

## 
## [[10]]

## 
## [[11]]

plt <- plots[[1]]
for (pi in 2:length(plots)) {
    plt <- plt + plots[[pi]]
}
plt <- plt + plot_layout(nrow = 4, ncol = 3)
plt

# ggsave("../campaign-adgroup/ukraine/figures/adgroup_reach_activity.png", plt, dpi = 300, bg = 'white', width = 13, height = 8)

tempdt <- select(d0, starts_with("subcamp"))
c <- correlation(tempdt, p_adjust = "none")
c

## # Correlation Matrix (pearson-method)
## 
## Parameter1       |       Parameter2 |    r |       95% CI | t(17047) |         p
## --------------------------------------------------------------------------------
## subcamp01_tweets | subcamp02_tweets | 0.83 | [0.82, 0.83] |   192.55 | < .001***
## subcamp01_tweets | subcamp03_tweets | 0.57 | [0.56, 0.58] |    89.82 | < .001***
## subcamp01_tweets | subcamp04_tweets | 0.69 | [0.68, 0.69] |   123.29 | < .001***
## subcamp01_tweets | subcamp05_tweets | 0.76 | [0.75, 0.76] |   151.02 | < .001***
## subcamp02_tweets | subcamp03_tweets | 0.61 | [0.60, 0.62] |    99.38 | < .001***
## subcamp02_tweets | subcamp04_tweets | 0.72 | [0.71, 0.73] |   134.76 | < .001***
## subcamp02_tweets | subcamp05_tweets | 0.80 | [0.79, 0.80] |   172.31 | < .001***
## subcamp03_tweets | subcamp04_tweets | 0.73 | [0.72, 0.73] |   138.52 | < .001***
## subcamp03_tweets | subcamp05_tweets | 0.58 | [0.57, 0.59] |    92.02 | < .001***
## subcamp04_tweets | subcamp05_tweets | 0.71 | [0.70, 0.71] |   130.61 | < .001***
## 
## p-value adjustment method: none
## Observations: 17049

c2 <- correlation(tempdt, p_adjust = "none", method = "kendall")
c2

## # Correlation Matrix (kendall-method)
## 
## Parameter1       |       Parameter2 |  tau |       95% CI |      z |         p
## ------------------------------------------------------------------------------
## subcamp01_tweets | subcamp02_tweets | 0.67 | [0.67, 0.68] | 126.84 | < .001***
## subcamp01_tweets | subcamp03_tweets | 0.62 | [0.61, 0.63] | 116.79 | < .001***
## subcamp01_tweets | subcamp04_tweets | 0.66 | [0.65, 0.66] | 124.07 | < .001***
## subcamp01_tweets | subcamp05_tweets | 0.63 | [0.62, 0.63] | 118.08 | < .001***
## subcamp02_tweets | subcamp03_tweets | 0.68 | [0.67, 0.68] | 128.13 | < .001***
## subcamp02_tweets | subcamp04_tweets | 0.69 | [0.69, 0.70] | 131.32 | < .001***
## subcamp02_tweets | subcamp05_tweets | 0.68 | [0.67, 0.68] | 127.76 | < .001***
## subcamp03_tweets | subcamp04_tweets | 0.71 | [0.71, 0.72] | 135.65 | < .001***
## subcamp03_tweets | subcamp05_tweets | 0.67 | [0.67, 0.68] | 126.98 | < .001***
## subcamp04_tweets | subcamp05_tweets | 0.71 | [0.71, 0.72] | 135.80 | < .001***
## 
## p-value adjustment method: none
## Observations: 17049

tempplt <- ggpairs(tempdt)
# ggsave("../campaign-adgroup/ukraine/figures/adgroup_activity_cor_user.png", tempplt, dpi = 300, bg = 'white', width = 13, height = 8)

tempdt <- select(d0, starts_with("subcamp"), lineitem)
tempdt <- tempdt[, lapply(.SD, mean), lineitem]
tempdt

##      lineitem subcamp01_tweets subcamp02_tweets subcamp03_tweets
##  1:  adgroup7         44.31912         59.88551         56.09622
##  2: adgroup11         14.17311         16.94843         16.76427
##  3: adgroup10         53.46053         64.48684         54.37719
##  4: adgroup23         79.72023         97.51506         89.21951
##  5:  adgroup9         35.30556         46.09259         45.10053
##  6:  adgroup6         42.98381         56.07490         56.24291
##  7: adgroup19         46.64358         58.58438         64.81864
##  8: adgroup17         82.59644         98.67507         93.86944
##  9:  adgroup8         77.14056        101.97323         97.62115
## 10: adgroup18         51.90168         62.42432         60.11125
## 11:  adgroup1         46.91933         55.70932         62.47844
## 12:  adgroup3         60.84332         71.23569         74.49591
## 13:  adgroup2         50.78182         65.55844         65.47273
## 14: adgroup25         32.60794         40.88337         39.53598
## 15: adgroup21         52.71543         66.34707         72.10904
## 16:  adgroup4         72.12639         87.75166         93.70953
## 17: adgroup13        104.14060        130.43042        122.40316
## 18: adgroup15         84.64831        105.51189         85.68335
## 19: adgroup16         75.72397        101.73884         95.90744
## 20:  adgroup0         49.34364         59.76804         57.11512
## 21:  adgroup5         36.86469         50.09076         47.52310
## 22: adgroup26        118.85748        142.76960        125.98812
## 23: adgroup22         65.44420         77.27133         76.88403
## 24: adgroup12         85.14147        104.82108        102.60055
## 25: adgroup24         62.17808         74.03253         70.80479
## 26: adgroup20        166.69575        205.60403        199.55928
## 27: adgroup14         82.54945        101.69963        101.82234
##      lineitem subcamp01_tweets subcamp02_tweets subcamp03_tweets
##     subcamp04_tweets subcamp05_tweets
##  1:         98.25213         63.91474
##  2:         28.32781         17.53407
##  3:         97.10526         59.48026
##  4:        164.60115        106.96987
##  5:         73.18122         51.28439
##  6:         99.77126         57.49393
##  7:        100.81864         64.75063
##  8:        161.00148        119.58012
##  9:        157.20750        101.11647
## 10:         98.86805         69.05433
## 11:         93.39360         62.09040
## 12:        111.65804         82.52452
## 13:        110.48571         68.01688
## 14:         65.97767         37.67494
## 15:        112.72074         68.31516
## 16:        135.57871        100.47450
## 17:        188.55524        140.91535
## 18:        168.11264        110.84481
## 19:        175.98843        106.88760
## 20:        104.09450         60.93471
## 21:         83.59736         52.86469
## 22:        222.34917        152.71021
## 23:        149.92341         82.74617
## 24:        175.38696        112.74619
## 25:        124.74315         76.18664
## 26:        354.80761        222.20134
## 27:        163.17399        111.72344
##     subcamp04_tweets subcamp05_tweets

tempdt[, lineitem := NULL]
c <- correlation(tempdt, p_adjust = "none")
c

## # Correlation Matrix (pearson-method)
## 
## Parameter1       |       Parameter2 |    r |       95% CI | t(25) |         p
## -----------------------------------------------------------------------------
## subcamp01_tweets | subcamp02_tweets | 1.00 | [0.99, 1.00] | 59.53 | < .001***
## subcamp01_tweets | subcamp03_tweets | 0.98 | [0.97, 0.99] | 28.45 | < .001***
## subcamp01_tweets | subcamp04_tweets | 0.99 | [0.97, 0.99] | 29.10 | < .001***
## subcamp01_tweets | subcamp05_tweets | 0.99 | [0.99, 1.00] | 45.92 | < .001***
## subcamp02_tweets | subcamp03_tweets | 0.99 | [0.97, 0.99] | 31.32 | < .001***
## subcamp02_tweets | subcamp04_tweets | 0.99 | [0.98, 0.99] | 33.03 | < .001***
## subcamp02_tweets | subcamp05_tweets | 0.99 | [0.99, 1.00] | 45.26 | < .001***
## subcamp03_tweets | subcamp04_tweets | 0.98 | [0.96, 0.99] | 27.02 | < .001***
## subcamp03_tweets | subcamp05_tweets | 0.99 | [0.97, 0.99] | 32.04 | < .001***
## subcamp04_tweets | subcamp05_tweets | 0.98 | [0.96, 0.99] | 26.76 | < .001***
## 
## p-value adjustment method: none
## Observations: 27

c2 <- correlation(tempdt, p_adjust = "none", method = "kendall")
c2

## # Correlation Matrix (kendall-method)
## 
## Parameter1       |       Parameter2 |  tau |       95% CI |    z |         p
## ----------------------------------------------------------------------------
## subcamp01_tweets | subcamp02_tweets | 0.91 | [0.85, 0.95] | 6.65 | < .001***
## subcamp01_tweets | subcamp03_tweets | 0.83 | [0.73, 0.90] | 6.11 | < .001***
## subcamp01_tweets | subcamp04_tweets | 0.85 | [0.76, 0.91] | 6.23 | < .001***
## subcamp01_tweets | subcamp05_tweets | 0.90 | [0.83, 0.94] | 6.57 | < .001***
## subcamp02_tweets | subcamp03_tweets | 0.85 | [0.75, 0.91] | 6.19 | < .001***
## subcamp02_tweets | subcamp04_tweets | 0.87 | [0.79, 0.93] | 6.40 | < .001***
## subcamp02_tweets | subcamp05_tweets | 0.86 | [0.78, 0.92] | 6.32 | < .001***
## subcamp03_tweets | subcamp04_tweets | 0.86 | [0.77, 0.91] | 6.27 | < .001***
## subcamp03_tweets | subcamp05_tweets | 0.88 | [0.80, 0.93] | 6.44 | < .001***
## subcamp04_tweets | subcamp05_tweets | 0.85 | [0.76, 0.91] | 6.23 | < .001***
## 
## p-value adjustment method: none
## Observations: 27

tempplt <- ggpairs(tempdt)
# ggsave("../campaign-adgroup/ukraine/figures/adgroup_activity_cor_adgroup.png", tempplt, dpi = 300, bg = 'white', width = 13, height = 8)

tempdt <- data.table()
c <- 1
for (c in 1:length(campaigns)) {
    camp <- campaigns[c]
    tempdat <- d3[campaign == camp & condition == "t"]
    temp_reach <- fread(reach[c])
    if (c >= 6){
      temp_reach <- temp_reach %>% filter(Objective == "Video views") %>% filter(`Ad Group status` == "Running")
      temp_reach$`Ad Group name` <- sapply(temp_reach$`Ad Group name`, function(x) change_ad_group_name(x))
    }
    temp_reach <- temp_reach[, .(`Ad Group name`, `Total audience reach`)]
    setnames(temp_reach, c("lineitem", "reach"))
    tempdat <- left_join(tempdat, temp_reach)
    setDT(tempdat)
    tempdat[, reach := as.numeric(reach)]
    tempdat$reach[is.na(tempdat$reach)] <- 0
    tempdat[, reach_prop := reach / N]
    tempdat <- tempdat[, .(lineitem, campaign, N, reach, reach_prop)]
    temp_name <- as.vector(tempdat[, unique(campaign)])
    tempdat <- tempdat[, .(lineitem, reach_prop)]
    setnames(tempdat, "reach_prop", temp_name)
    if (nrow(tempdt) == 0) {
        tempdt <- tempdat
    }
    tempdt <- left_join(tempdt, tempdat) |> data.table()
    tempdt
}

## Joining, by = "lineitem"

## Joining, by = c("lineitem", "subcamp01_tweets")

## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"

tempdt

##      lineitem subcamp01_tweets subcamp02_tweets subcamp03_tweets
##  1:  adgroup0       0.10616438       0.22945205       0.18835616
##  2:  adgroup1       0.07022472       0.20505618       0.15730337
##  3:  adgroup2       0.10966057       0.20626632       0.17754569
##  4:  adgroup3       0.01928375       0.05785124       0.07713499
##  5:  adgroup4       0.01339286       0.06696429       0.07589286
##  6:  adgroup5       0.15946844       0.34219269       0.31229236
##  7:  adgroup6       0.09465021       0.22633745       0.22222222
##  8:  adgroup7       0.04645477       0.13447433       0.12224939
##  9:  adgroup8       0.11081081       0.22162162       0.19729730
## 10:  adgroup9       0.08266667       0.17333333       0.16800000
## 11: adgroup10       0.09170306       0.21397380       0.20087336
## 12: adgroup11       0.04832714       0.13754647       0.11895911
## 13: adgroup12       0.07777778       0.16111111       0.14166667
## 14: adgroup13       0.01719198       0.02865330       0.04871060
## 15: adgroup14       0.02545455       0.06909091       0.08000000
## 16: adgroup15       0.03990025       0.16957606       0.13216958
## 17: adgroup16       0.04333333       0.17000000       0.18333333
## 18: adgroup17       0.05029586       0.12426036       0.14792899
## 19: adgroup18       0.04392765       0.20930233       0.18087855
## 20: adgroup19       0.08883249       0.25380711       0.21065990
## 21: adgroup20       0.04845815       0.12775330       0.13656388
## 22: adgroup21       0.10928962       0.23770492       0.20218579
## 23: adgroup22       0.15948276       0.32758621       0.25862069
## 24: adgroup23       0.06956522       0.19130435       0.15072464
## 25: adgroup24       0.07823129       0.22448980       0.20068027
## 26: adgroup25       0.07035176       0.15075377       0.16582915
## 27: adgroup26       0.07729469       0.14975845       0.12077295
##      lineitem subcamp01_tweets subcamp02_tweets subcamp03_tweets
##     subcamp04_tweets subcamp05_tweets subcamp06_tweets subcamp07_tweets
##  1:       0.19863014       0.19178082       0.17465753       0.17123288
##  2:       0.19101124       0.16853933       0.12078652       0.12921348
##  3:       0.19321149       0.19582245       0.15665796       0.16449086
##  4:       0.06336088       0.07713499       0.04407713       0.03856749
##  5:       0.07589286       0.04910714       0.04464286       0.03125000
##  6:       0.38538206       0.33887043       0.33887043       0.26245847
##  7:       0.24279835       0.29218107       0.21810700       0.22633745
##  8:       0.13447433       0.11735941       0.10757946       0.11735941
##  9:       0.23783784       0.21351351       0.19729730       0.16216216
## 10:       0.18400000       0.18133333       0.12800000       0.13866667
## 11:       0.23144105       0.19650655       0.19213974       0.18340611
## 12:       0.13754647       0.12267658       0.08921933       0.10037175
## 13:       0.15277778       0.16666667       0.15833333       0.12777778
## 14:       0.04011461       0.05444126       0.02578797       0.02865330
## 15:       0.08363636       0.05818182       0.04363636       0.04363636
## 16:       0.16708229       0.15710723       0.14962594       0.15461347
## 17:       0.20333333       0.15333333       0.13666667       0.17333333
## 18:       0.13609467       0.16272189       0.11538462       0.13905325
## 19:       0.18087855       0.18087855       0.14728682       0.15503876
## 20:       0.25634518       0.21573604       0.20558376       0.19543147
## 21:       0.15859031       0.14977974       0.13215859       0.08810573
## 22:       0.20765027       0.18306011       0.19125683       0.18032787
## 23:       0.31465517       0.24137931       0.22844828       0.25431034
## 24:       0.15942029       0.19420290       0.14492754       0.15942029
## 25:       0.17006803       0.16666667       0.18367347       0.14285714
## 26:       0.16080402       0.13065327       0.11055276       0.13065327
## 27:       0.16908213       0.14492754       0.13526570       0.12077295
##     subcamp04_tweets subcamp05_tweets subcamp06_tweets subcamp07_tweets
##     subcamp08_tweets subcamp09_tweets subcamp10_tweets subcamp11_tweets
##  1:       0.16095890       0.14383562       0.15410959       0.17465753
##  2:       0.18820225       0.14044944       0.11235955       0.14887640
##  3:       0.14882507       0.13838120       0.13054830       0.14621410
##  4:       0.03856749       0.04683196       0.03856749       0.04958678
##  5:       0.06250000       0.05357143       0.07589286       0.03571429
##  6:       0.22259136       0.26245847       0.20598007       0.28239203
##  7:       0.22633745       0.21810700       0.17283951       0.23045267
##  8:       0.10024450       0.09046455       0.08068460       0.10024450
##  9:       0.15675676       0.18648649       0.12702703       0.14324324
## 10:       0.13066667       0.14933333       0.12533333       0.15200000
## 11:       0.18340611       0.20087336       0.15720524       0.17030568
## 12:       0.11152416       0.09665428       0.07063197       0.09293680
## 13:       0.11111111       0.12500000       0.09722222       0.13055556
## 14:       0.04297994       0.03438395       0.03151862       0.02292264
## 15:       0.05090909       0.06545455       0.03272727       0.04000000
## 16:       0.13965087       0.13216958       0.09226933       0.14713217
## 17:       0.17000000       0.18333333       0.13666667       0.13000000
## 18:       0.12426036       0.12130178       0.08579882       0.07988166
## 19:       0.16795866       0.14211886       0.14728682       0.14987080
## 20:       0.15228426       0.16497462       0.12690355       0.13959391
## 21:       0.09691630       0.11013216       0.08370044       0.07488987
## 22:       0.15300546       0.15846995       0.16120219       0.18032787
## 23:       0.20689655       0.18103448       0.20258621       0.23706897
## 24:       0.12463768       0.12173913       0.10724638       0.14202899
## 25:       0.13945578       0.13945578       0.12585034       0.16666667
## 26:       0.12060302       0.11557789       0.10552764       0.11557789
## 27:       0.08212560       0.13526570       0.07246377       0.11594203
##     subcamp08_tweets subcamp09_tweets subcamp10_tweets subcamp11_tweets

tempdt[, lineitem := NULL]

c <- correlation(tempdt, p_adjust = "none")
c

## # Correlation Matrix (pearson-method)
## 
## Parameter1       |       Parameter2 |    r |       95% CI | t(25) |         p
## -----------------------------------------------------------------------------
## subcamp01_tweets | subcamp02_tweets | 0.91 | [0.80, 0.96] | 10.75 | < .001***
## subcamp01_tweets | subcamp03_tweets | 0.88 | [0.76, 0.95] |  9.46 | < .001***
## subcamp01_tweets | subcamp04_tweets | 0.89 | [0.78, 0.95] |  9.96 | < .001***
## subcamp01_tweets | subcamp05_tweets | 0.85 | [0.70, 0.93] |  8.22 | < .001***
## subcamp02_tweets | subcamp03_tweets | 0.96 | [0.92, 0.98] | 17.46 | < .001***
## subcamp02_tweets | subcamp04_tweets | 0.95 | [0.90, 0.98] | 15.82 | < .001***
## subcamp02_tweets | subcamp05_tweets | 0.91 | [0.81, 0.96] | 10.90 | < .001***
## subcamp03_tweets | subcamp04_tweets | 0.96 | [0.92, 0.98] | 18.07 | < .001***
## subcamp03_tweets | subcamp05_tweets | 0.93 | [0.85, 0.97] | 12.72 | < .001***
## subcamp04_tweets | subcamp05_tweets | 0.93 | [0.85, 0.97] | 12.61 | < .001***
## 
## p-value adjustment method: none
## Observations: 27

c2 <- correlation(tempdt, p_adjust = "none", method = "kendall")
c2

## # Correlation Matrix (kendall-method)
## 
## Parameter1       |       Parameter2 |  tau |       95% CI |    z |         p
## ----------------------------------------------------------------------------
## subcamp01_tweets | subcamp02_tweets | 0.66 | [0.48, 0.79] | 4.86 | < .001***
## subcamp01_tweets | subcamp03_tweets | 0.67 | [0.49, 0.79] | 4.90 | < .001***
## subcamp01_tweets | subcamp04_tweets | 0.66 | [0.48, 0.79] | 4.86 | < .001***
## subcamp01_tweets | subcamp05_tweets | 0.70 | [0.54, 0.81] | 5.13 | < .001***
## subcamp02_tweets | subcamp03_tweets | 0.81 | [0.70, 0.89] | 5.94 | < .001***
## subcamp02_tweets | subcamp04_tweets | 0.81 | [0.69, 0.88] | 5.90 | < .001***
## subcamp02_tweets | subcamp05_tweets | 0.75 | [0.61, 0.85] | 5.50 | < .001***
## subcamp03_tweets | subcamp04_tweets | 0.80 | [0.68, 0.88] | 5.86 | < .001***
## subcamp03_tweets | subcamp05_tweets | 0.76 | [0.62, 0.85] | 5.55 | < .001***
## subcamp04_tweets | subcamp05_tweets | 0.76 | [0.63, 0.86] | 5.59 | < .001***
## 
## p-value adjustment method: none
## Observations: 27

tempplt <- ggpairs(tempdt)
# ggsave("../campaign-adgroup/ukraine/figures/adgroup_reach_cor_adgroup.png", tempplt, dpi = 300, bg = 'white', width = 13, height = 8)

Scatterplot

Reach Prop vs Mean Number of Tweets

plots <- list()
full_dat <- c()
for (c in 1:length(campaigns)) {
    camp <- campaigns[c]
    tempdat <- d3[campaign == camp & condition == "t"]
    temp_reach <- fread(reach[c])
    if (c >= 6){
      temp_reach <- temp_reach %>% filter(Objective == "Video views") %>% filter(`Ad Group status` == "Running")
      temp_reach$`Ad Group name` <- sapply(temp_reach$`Ad Group name`, function(x) change_ad_group_name(x))
    }
    temp_reach <- temp_reach[, .(`Ad Group name`, `Total audience reach`)]
    setnames(temp_reach, c("lineitem", "reach"))
    tempdat <- left_join(tempdat, temp_reach)
    setDT(tempdat)
    tempdat[, reach := as.numeric(reach)]
    tempdat$reach[is.na(tempdat$reach)] <- 0
    tempdat[, reach_prop := reach / N]
    ord <- tempdat[order(reach_prop), lineitem]
    tempdat[, lineitem := factor(lineitem, levels = ord)]
    full_dat <- rbind(full_dat, tempdat)
    
    p1 <- ggplot(tempdat, aes(x = reach_prop, y = avg_active_tweets)) +
        geom_point(aes(color = lineitem)) +
        # geom_text(aes(label = paste0(active_user_count, "/", N)), size = 3) +
        scale_x_continuous(limits = c(0, 0.5)) +
        labs(x = "proportion reach (treatment)", y = "mean number of tweets", title = glue("day {c}"))
    p1
    plots[[c]] <- p1
}

## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"

plots

## [[1]]

## 
## [[2]]

## 
## [[3]]

## 
## [[4]]

## 
## [[5]]

## 
## [[6]]

## 
## [[7]]

## 
## [[8]]

## 
## [[9]]

## 
## [[10]]

## 
## [[11]]

full_dat %>% group_by(campaign) %>% summarize(cor = cor(reach_prop, avg_active_tweets))

## # A tibble: 11 × 2
##    campaign            cor
##    <fct>             <dbl>
##  1 subcamp01_tweets -0.361
##  2 subcamp02_tweets -0.393
##  3 subcamp03_tweets -0.390
##  4 subcamp04_tweets -0.237
##  5 subcamp05_tweets -0.350
##  6 subcamp06_tweets -0.240
##  7 subcamp07_tweets -0.261
##  8 subcamp08_tweets -0.321
##  9 subcamp09_tweets -0.307
## 10 subcamp10_tweets -0.325
## 11 subcamp11_tweets -0.327

Reach Prop vs Super User Prop

plots <- list()
full_dat <- c()
for (c in 1:length(campaigns)) {
    camp <- campaigns[c]
    tempdat <- d3[campaign == camp & condition == "t"]
    temp_reach <- fread(reach[c])
    if (c >= 6){
      temp_reach <- temp_reach %>% filter(Objective == "Video views") %>% filter(`Ad Group status` == "Running")
      temp_reach$`Ad Group name` <- sapply(temp_reach$`Ad Group name`, function(x) change_ad_group_name(x))
    }
    temp_reach <- temp_reach[, .(`Ad Group name`, `Total audience reach`)]
    setnames(temp_reach, c("lineitem", "reach"))
    tempdat <- left_join(tempdat, temp_reach)
    setDT(tempdat)
    tempdat[, reach := as.numeric(reach)]
    tempdat$reach[is.na(tempdat$reach)] <- 0
    tempdat[, reach_prop := reach / N]
    ord <- tempdat[order(reach_prop), lineitem]
    tempdat[, lineitem := factor(lineitem, levels = ord)]
    full_dat <- rbind(full_dat, tempdat)
    
    p1 <- ggplot(tempdat, aes(x = reach_prop, y = super_user_prop)) +
        geom_point(aes(color = lineitem)) +
        # geom_text(aes(label = paste0(active_user_count, "/", N)), size = 3) +
        scale_x_continuous(limits = c(0, 0.5)) +
        labs(x = "proportion reach (treatment)", y = "proportion of super users", title = glue("day {c}"))
    p1
    plots[[c]] <- p1
}

## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"

plots

## [[1]]

## 
## [[2]]

## 
## [[3]]

## 
## [[4]]

## 
## [[5]]

## 
## [[6]]

## 
## [[7]]

## 
## [[8]]

## 
## [[9]]

## 
## [[10]]

## 
## [[11]]

full_dat %>% group_by(campaign) %>% summarise(cor = cor(reach_prop, super_user_prop))

## # A tibble: 11 × 2
##    campaign             cor
##    <fct>              <dbl>
##  1 subcamp01_tweets -0.391 
##  2 subcamp02_tweets -0.419 
##  3 subcamp03_tweets -0.446 
##  4 subcamp04_tweets -0.201 
##  5 subcamp05_tweets -0.427 
##  6 subcamp06_tweets -0.124 
##  7 subcamp07_tweets -0.0294
##  8 subcamp08_tweets -0.172 
##  9 subcamp09_tweets -0.344 
## 10 subcamp10_tweets -0.262 
## 11 subcamp11_tweets -0.286

Reach Prop vs Idle User Prop

plots <- list()
full_dat <- c()
for (c in 1:length(campaigns)) {
    camp <- campaigns[c]
    tempdat <- d3[campaign == camp & condition == "t"]
    temp_reach <- fread(reach[c])
    if (c >= 6){
      temp_reach <- temp_reach %>% filter(Objective == "Video views") %>% filter(`Ad Group status` == "Running")
      temp_reach$`Ad Group name` <- sapply(temp_reach$`Ad Group name`, function(x) change_ad_group_name(x))
    }
    temp_reach <- temp_reach[, .(`Ad Group name`, `Total audience reach`)]
    setnames(temp_reach, c("lineitem", "reach"))
    tempdat <- left_join(tempdat, temp_reach)
    setDT(tempdat)
    tempdat[, reach := as.numeric(reach)]
    tempdat$reach[is.na(tempdat$reach)] <- 0
    tempdat[, reach_prop := reach / N]
    ord <- tempdat[order(reach_prop), lineitem]
    tempdat[, lineitem := factor(lineitem, levels = ord)]
    full_dat <- rbind(full_dat, tempdat)
    
    p1 <- ggplot(tempdat, aes(x = reach_prop, y = idle_user_prop)) +
        geom_point(aes(color = lineitem)) +
        # geom_text(aes(label = paste0(active_user_count, "/", N)), size = 3) +
        scale_x_continuous(limits = c(0, 0.5)) +
        labs(x = "proportion reach (treatment)", y = "proportion of idle users", title = glue("day {c}"))
    p1
    plots[[c]] <- p1
}

## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"
## Joining, by = "lineitem"

plots

## [[1]]

## 
## [[2]]

## 
## [[3]]

## 
## [[4]]

## 
## [[5]]

## 
## [[6]]

## 
## [[7]]

## 
## [[8]]

## 
## [[9]]

## 
## [[10]]

## 
## [[11]]

full_dat %>% group_by(campaign) %>% summarise(cor = cor(reach_prop, idle_user_prop))

## # A tibble: 11 × 2
##    campaign           cor
##    <fct>            <dbl>
##  1 subcamp01_tweets 0.155
##  2 subcamp02_tweets 0.346
##  3 subcamp03_tweets 0.266
##  4 subcamp04_tweets 0.146
##  5 subcamp05_tweets 0.259
##  6 subcamp06_tweets 0.210
##  7 subcamp07_tweets 0.283
##  8 subcamp08_tweets 0.163
##  9 subcamp09_tweets 0.495
## 10 subcamp10_tweets 0.407
## 11 subcamp11_tweets 0.350