Using dplyr's summarise on single column, but with multiple parameter values - r

Question

Apologize for the not-so-clear title (could use help) - hopefully the example below will clarify many things. I have the following dataframe of basketball shot results (1 row == 1 basketball shot):
> dput(zed)
structure(list(shooterTeamAlias = c("DUKE", "DUKE", "BC", "DUKE",
"DUKE", "DUKE", "DUKE", "DUKE", "DUKE", "BC", "BC", "BC", "DUKE",
"BC", "BC", "DUKE", "DUKE", "DUKE", "BC", "DUKE"), distanceCategory = c("sht2",
"sht2", "sht3", "atr2", "mid2", "sht2", "lng3", "sht3", "atr2",
"sht3", "sht3", "sht2", "mid2", "sht3", "sht3", "sht3", "atr2",
"atr2", "sht2", "mid2"), eventType = c("twopointmiss", "twopointmade",
"threepointmade", "twopointmade", "twopointmiss", "twopointmade",
"threepointmiss", "threepointmiss", "twopointmade", "threepointmiss",
"threepointmade", "twopointmiss", "twopointmade", "threepointmiss",
"threepointmade", "threepointmiss", "twopointmade", "twopointmade",
"twopointmade", "twopointmade")), row.names = c(NA, 20L), class = "data.frame")
> zed
shooterTeamAlias distanceCategory eventType
1 DUKE sht2 twopointmiss
2 DUKE sht2 twopointmade
3 BC sht3 threepointmade
4 DUKE atr2 twopointmade
5 DUKE mid2 twopointmiss
6 DUKE sht2 twopointmade
7 DUKE lng3 threepointmiss
8 DUKE sht3 threepointmiss
9 DUKE atr2 twopointmade
10 BC sht3 threepointmiss
11 BC sht3 threepointmade
12 BC sht2 twopointmiss
13 DUKE mid2 twopointmade
14 BC sht3 threepointmiss
15 BC sht3 threepointmade
16 DUKE sht3 threepointmiss
17 DUKE atr2 twopointmade
18 DUKE atr2 twopointmade
19 BC sht2 twopointmade
20 DUKE mid2 twopointmade
This dataframe is currently in a tidy-ish format, and I need to group_by team and then fatten it big time. The full data has 6 distanceCategories atr2, sht2, mid2, lng2, sht3, lng3 (example above has 5 only), as well as 2 categories that are a function of the other 6: all2 is atr2, sht2, lng2, mid2 and all3 is sht3, lng3. For each of these 8 categories then, I would like a column for makes, attempts, pct, and attempt frequency. I use the eventType column to determine if a shot was made. I am currently doing so with the following
fat.data <- {zed %>%
dplyr::group_by(shooterTeamAlias) %>%
dplyr::summarise(
shotsCount = n(),
# Shooting By Distance Stats
atr2Made = sum(distanceCategory == "atr2" & eventType == "twopointmade"),
atr2Att = sum(distanceCategory == "atr2" & eventType %in% c("twopointmiss", "twopointmade")),
atr2AttFreq = atr2Att / shotsCount,
atr2Pct = ifelse(atr2Att > 0, atr2Made / atr2Att, 0),
sht2Made = sum(distanceCategory == "sht2" & eventType == "twopointmade"),
sht2Att = sum(distanceCategory == "sht2" & eventType %in% c("twopointmiss", "twopointmade")),
sht2AttFreq = sht2Att / shotsCount,
sht2Pct = ifelse(sht2Att > 0, sht2Made / sht2Att, 0),
mid2Made = sum(distanceCategory == "mid2" & eventType == "twopointmade"),
mid2Att = sum(distanceCategory == "mid2" & eventType %in% c("twopointmiss", "twopointmade")),
mid2AttFreq = mid2Att / shotsCount,
mid2Pct = ifelse(mid2Att > 0, mid2Made / mid2Att, 0),
lng2Made = sum(distanceCategory == "lng2" & eventType == "twopointmade"),
lng2Att = sum(distanceCategory == "lng2" & eventType %in% c("twopointmiss", "twopointmade")),
lng2AttFreq = lng2Att / shotsCount,
lng2Pct = ifelse(lng2Att > 0, lng2Made / lng2Att, 0),
all2Made = sum(atr2Made, sht2Made, mid2Made, lng2Made),
all2Att = sum(atr2Att, sht2Att, mid2Att, lng2Att),
all2AttFreq = all2Att / shotsCount,
all2Pct = ifelse(all2Att > 0, all2Made / all2Att, 0),
sht3Made = sum(distanceCategory == "sht3" & eventType == "threepointmade"),
sht3Att = sum(distanceCategory == "sht3" & eventType %in% c("threepointmiss", "threepointmade")),
sht3AttFreq = sht3Att / shotsCount,
sht3Pct = ifelse(sht3Att > 0, sht3Made / sht3Att, 0),
lng3Made = sum(distanceCategory == "lng3" & eventType == "threepointmade"),
lng3Att = sum(distanceCategory == "lng3" & eventType %in% c("threepointmiss", "threepointmade")),
lng3AttFreq = lng3Att / shotsCount,
lng3Pct = ifelse(lng3Att > 0, lng3Made / lng3Att, 0),
all3Made = sum(sht3Made, lng3Made),
all3Att = sum(sht3Att, lng3Att),
all3AttFreq = all3Att / shotsCount,
all3Pct = ifelse(all3Att > 0, all3Made / all3Att, 0))}
...for the 6 categories that appear in the data (all but all2 and all3), their 4 columns are all computed in the same manner. As you'll see for all2 and all3, the calculations are a bit different.
Not worrying for the time being about the all2 and all3 categories, is there a better way to compute the makes, attempts, pct, and attempt frequencies for the 6 categories in the data? For the 8 categories * 4 column-types == 32 columns here, it's not so bad, but I have another, similar instance where I have 21 categories * 4 column-types, and I have to do this multiple times in my code.
Not sure if dplyr::group_by dplyr::summarise is my best option (obv it's what im using currently), or if there's a better way to go about this. Improving this code / potentially speeding it up for my project is pivotally important, and any help is appreciated / i'll try to remember to bounty this post even if answered in the next 2 days.
Edit !!! : I've just realized that grouping by the distanceCategory first, computing the 4 stats for each distanceCategory, and then re-structuring that dataframe into this fat format may be easier... it is something I'm working on computing currently. Something along these lines:
zed %>%
dplyr::group_by(shooterTeamAlias, distanceCategory) %>%
dplyr::summarise(
attempts = ...,
makes = ...,
pct = ...,
attfreq = ...
) %>%
tidyr::spread(...)
Thanks!!

Answer 1

This looks like it could be made simpler by grouping by distanceCategory and then applying the same logic to each:
library(tidyverse)
zed %>%
group_by(shooterTeamAlias, distanceCategory) %>%
summarize(att = n(), # n() counts how many rows in this group
made = sum(eventType %>% str_detect("made"))
pct = if_else(att > 0, made / att, 0)) %>%
mutate(freq = att / sum(att))
# A tibble: 7 x 6
# Groups: shooterTeamAlias [2]
shooterTeamAlias distanceCategory att made pct freq
<chr> <chr> <int> <int> <dbl> <dbl>
1 BC sht2 2 1 0.5 0.286
2 BC sht3 5 3 0.6 0.714
3 DUKE atr2 4 4 1 0.308
4 DUKE lng3 1 0 0 0.0769
5 DUKE mid2 3 2 0.667 0.231
6 DUKE sht2 3 2 0.667 0.231
7 DUKE sht3 2 0 0 0.154
If you want that in wide format, you could first gather the calculations above, unite the distance with the stat, and then spread by that:
[same code as above] %>%
gather(stat, value, -distanceCategory, -shooterTeamAlias) %>%
unite(stat, distanceCategory, stat) %>%
spread(stat, value)
# A tibble: 2 x 21
# Groups: shooterTeamAlias [2]
shooterTeamAlias atr2_att atr2_freq atr2_made atr2_pct lng3_att lng3_freq lng3_made lng3_pct mid2_att mid2_freq mid2_made mid2_pct sht2_att sht2_freq sht2_made sht2_pct sht3_att sht3_freq sht3_made sht3_pct
<chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 BC NA NA NA NA NA NA NA NA NA NA NA NA 2 0.286 1 0.5 5 0.714 3 0.6
2 DUKE 4 0.308 4 1 1 0.0769 0 0 3 0.231 2 0.667 3 0.231 2 0.667 2 0.154 0 0