oa_exploration.Rmd

---
output: github_document
---

```{r, setup, echo=FALSE, message = FALSE}
knitr::opts_chunk$set(
  comment = "#>",
  collapse = TRUE,
  warning = FALSE,
  message = FALSE,
  echo = TRUE,
  fig.width = 6,
  fig.asp = 0.618,
  out.width = "70%",
  fig.align = "center",
  dpi = 300
)
options(scipen = 999, digits = 4)
knitr::knit_hooks$set(
  inline = function(x) {
    if (is.numeric(x)) {
      return(prettyNum(x, big.mark = ","))
    } else{
      return(x)
    }
  }
)
library(tidyverse)
library(jsonlite)
```

## Data

Patents matched to Unpaywall via DOIs

```{r}
npl_df <- readr::read_csv("data/dois_to_be_checked.csv") %>%
  mutate(doi = tolower(doi_cleaned))

oa_upy <- jsonlite::stream_in(file("data/oa_patents.json"), verbose = FALSE)
my_df <-
  inner_join(npl_df, oa_upy, by = "doi")
my_df
```

## Exploratory data analysis

### Publication view

#### By publication type

```{r}
my_df %>%
  group_by(genre) %>%
  summarise(n = n_distinct(doi)) %>%
  mutate(prop = n / sum(n)) %>%
  arrange(desc(n)) %>%
  knitr::kable()
```

#### By publication year (journal articles only)

```{r}
my_df %>%
  filter(genre == "journal-article") %>%
  group_by(year, is_oa) %>%
  summarise(n = n_distinct(doi)) %>%
  ggplot(aes(year, n, fill = is_oa, group = is_oa)) +
  geom_area() +
   scale_fill_manual(
      values = c("#cccccca0", "#56b4e9"),
      name = NULL,
      labels = c("Closed", "Open Access")
    ) +
  scale_y_continuous(expand = expansion(mult = c(0, 0.05)), 
    #limits = c(0, 8000)
    ) +
  cowplot::theme_minimal_hgrid() +
  labs(x = "Publication Year", y = "NPLs") +
  theme(legend.position = "top",
          legend.justification = "right")
```

#### OA by provider

Unpaywall distinguishes the provider of the open access full text, `publisher` 
and `repository`.

```{r}
oa_df <- my_df %>%
  filter(genre == "journal-article", is_oa == TRUE) %>%
  unnest(oa_locations)

# overlap repo / publisher
host_cat <- oa_df %>%
  distinct(doi, year, host_type, is_best, has_repository_copy) %>%
  filter(is_best == TRUE) %>%
  mutate(provider_cat = case_when(
    host_type == "publisher" & has_repository_copy == TRUE ~ "Journal & Repository",
    host_type == "repository" ~ "Repository only",
    host_type == "publisher" ~ "Journal only"
  )) %>%
  group_by(provider_cat, year) %>%
  summarise(n = n_distinct(doi))

# all years
all_year <- oa_df %>%
  group_by(year) %>%
  summarise(n = n_distinct(doi))

ggplot(host_cat, aes(x = year, y = n)) +
  geom_bar(
    data = all_year,
    aes(fill = "All OA Articles"),
    color = "transparent",
    stat = "identity"
  ) +
  geom_bar(aes(fill = "by Host"), color = "transparent", stat = "identity") +
  facet_wrap( ~ provider_cat, nrow = 1) +
  scale_fill_manual(values = c("#b3b3b3a0", "#56B4E9"), name = "") +
  scale_y_continuous(
      labels = scales::number_format(big.mark = ","),
      expand = expansion(mult = c(0, 0.05))) +
  labs(x = "Publication Year", y = "OA Articles") +
  theme(legend.position = "top", legend.justification = "right") +
  cowplot::theme_minimal_hgrid() +
  theme(legend.position = "top",
           legend.justification = "right")
```

#### OA by Unpaywall OA color


```{r}
oa_df %>%
  group_by(oa_status) %>%
  summarise(n = n_distinct((doi))) %>%
  mutate(prop = n / length(unique(oa_df$doi))) %>% 
  arrange(desc(n)) %>%
  knitr::kable()
```

### Patent view

#### Lag patent and npl publication date

```{r}
my_df %>%
  filter(genre == "journal-article") %>%
  select(doi, publication_date, country_code, year) %>%
  mutate(npl_year = substr(publication_date, 1, 4)) %>%
  mutate(lag = as.integer(npl_year) - as.integer(year)) %>%
  ggplot(aes(x = lag)) +
  geom_histogram(stat = "count", fill="#56B4E9") +
  geom_vline(aes(xintercept = median(lag, na.rm = T)),
             colour = "#E69F00", linetype ="dashed", size = .8) +
  cowplot::theme_minimal_hgrid() +
  scale_y_continuous(expand = expand_scale(mult = c(0, 0.05))) +
  labs(y = "NPL", x = " Lag between patent issued year and npl publication year")
```

#### OA by patent office

Only US vs European Patent Office (EP)
```{r}
my_df %>%
  mutate(npl_year = substr(publication_date, 1, 4)) %>%
  # only us vs ep
  filter(country_code %in% c("US", "EP")) %>%
  group_by(is_oa, country_code, npl_year) %>%
  summarise(n = n_distinct(doi)) %>%
  ggplot(aes(as.integer(npl_year), n, fill = is_oa)) +
  geom_bar(stat = "identity") +
   scale_fill_manual(
      values = c("#cccccca0", "#56b4e9"),
      name = NULL,
      labels = c("Closed", "Open Access")
    ) +
  scale_x_continuous(limits = c(2008, 2022)) +
  scale_y_continuous(expand = expansion(mult = c(0, 0.05))) +
  facet_wrap(~country_code) +
  cowplot::theme_minimal_hgrid() +
  labs(x = "Patent Issued Year", y = "NPLs") +
  theme(legend.position = "top",
          legend.justification = "right")
```