Commit paper, code, data, and package.

.gitignore

@@ -0,0 +1,3 @@
+.Rproj.user
+.Rhistory
+.RData

LICENSE

@@ -1,6 +1,6 @@
 The MIT License (MIT)
 
-Copyright (c) 2015 Polynumeral
+Copyright (c) 2015 Academia.edu
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

README.md

@@ -0,0 +1,67 @@
+# Academia Citation Advantage Analysis
+
+The `acadcites` package contains the data and functions used in Niyazov, et. al. "Open Access Meets Discoverability: Citations to Articles Posted to Academia.edu."
+
+## Installing the R Package
+The easiest way to install the package and its depdendencies is by using `install_local` from the `devtools` package. (http://cran.r-project.org/web/packages/devtools/index.html)
+
+- Clone the repo:
+
+   ```{R}
+   git clone https://github.com/polynumeral/academia-citations
+   cd academia-citations
+   ```
+
+- From R:
+
+   ```{R}
+   install.packages('devtools') 
+   devtools::install_local('acadcites_0.1.tar.gz')
+   ```
+
+## Importing data
+The cleaned/combined dataset used for the analyses can be obtained by calling:
+
+```{R}
+library('acadcites')
+cites <- importData()
+```
+
+or just `cites <- acadcites::importData()` without the `library` import.
+
+## Reproducing tables from the article
+
+Tables from the article can be reproduced with the `makeTable` function.
+
+```{R}
+# Make Table 1 from the article.
+makeTable(2, cites)
+
+# |Journal                                                | # Articles| % Total|
+# |:------------------------------------------------------|----------:|-------:|
+# |Analytical Chemistry                                   |      1,537|   3.44%|
+# |PLoS One                                               |        492|   1.10%|
+# |Anesthesia and Analgesia                               |        430|   0.96%|
+# |Biological and Pharmaceutical Bulletin                 |        362|   0.81%|
+# |Analytical Methods: advancing methods and applications |        339|   0.76%|
+# |Analytical Biochemistry                                |        317|   0.71%|
+# |Applied Mechanics and Materials                        |        303|   0.68%|
+# |Bioconjugate Chemistry                                 |        299|   0.67%|
+# |Applied Physics Letters                                |        190|   0.43%|
+# |BioEssays                                              |        183|   0.41%|
+```
+
+
+## Reproducing figures from the article
+The `makeFigure` function reproduces figures from the article. Like `makeTable`,
+it takes a figure number and a citations data frame.
+
+```{r}
+makeFigure(1, cites)
+```
+
+
+## Package help
+
+See `help(package='acadcites')` for more help files on individual functions, or
+`vignette('acadcites')` for information similar to what's provided here.

acadcites/.Rbuildignore

@@ -0,0 +1,2 @@
+^.*\.Rproj$
+^\.Rproj\.user$

acadcites/.gitignore

@@ -0,0 +1,4 @@
+.Rproj.user
+.Rhistory
+.RData
+inst/doc

acadcites/DESCRIPTION

@@ -0,0 +1,24 @@
+Package: acadcites
+Title: Manage data and models to study effect of Academia.edu on citations.
+Version: 0.1
+Authors@R: "Carl Vogel <carl@polynumeral.com> [aut, cre]"
+Description: Manage data and run models to study the effect of posting to
+    Academia.edu on article citations.
+Depends:
+    R (>= 3.1.1)
+License: MIT
+LazyData: true
+Imports:
+    dplyr,
+    magrittr,
+    stringr,
+    reshape2,
+    ggplot2,
+    MASS,
+    pscl,
+    knitr,
+    scales,
+    stargazer,
+    memisc,
+    pander
+VignetteBuilder: knitr

acadcites/NAMESPACE

@@ -0,0 +1,29 @@
+exportPattern("^[^\\.]")
+import(dplyr)
+importFrom(stringr, str_detect)
+importFrom(stringr, str_trim)
+importFrom(stringr, str_replace_all)
+importFrom(stringr, str_replace)
+importFrom(magrittr, use_series)
+importFrom(magrittr, set_colnames)
+importFrom(magrittr, set_names)
+importFrom(magrittr, extract)
+importFrom(ggplot2, ggplot)
+importFrom(ggplot2, aes)
+importFrom(ggplot2, geom_boxplot)
+importFrom(ggplot2, geom_histogram)
+importFrom(ggplot2, geom_point)
+importFrom(ggplot2, stat_quantile)
+importFrom(ggplot2, facet_wrap)
+importFrom(ggplot2, labs)
+importFrom(ggplot2, xlim)
+importFrom(ggplot2, scale_x_continuous)
+importFrom(ggplot2, scale_y_continuous)
+importFrom(ggplot2, scale_colour_manual)
+importFrom(ggplot2, position_jitter)
+importFrom(ggplot2, theme_bw)
+importFrom(memisc, mtable)
+importFrom(memisc, relabel)
+S3method(getModelSummary, lm)
+S3method(getModelSummary, glm)
+S3method(getModelSummary, zeroinfl)

acadcites/R/bucket_analysis.R

@@ -0,0 +1,108 @@
+# Functions for comparing citations within Impact Factor buckets
+
+#' Group a variable into buckets based on its quantiles or those of another
+#' variable.
+#'
+#' @param x_quantile The variable to calculate quantile buckets from.
+#' @param x_bucket The variable to collect into the quantile buckets.
+#' @param nbuckets The number of quantile buckets to use. Specify this *or*
+#' `probs`, but not both.
+#' @param probs The vector of probabilities for the quantile bucket cut points.
+#' Specify this *or* `nbuckets`, but not both.
+#' @return A factor vector corresponding to `x_bucket` with bucket ranges.
+#' If an element of `x_bucket` is outside of the range of `x_quantile`, its
+#' bucket will be NA.
+#'
+quantileBuckets <- function(x_quantile, x_bucket=x_quantile, nbuckets=10, probs=NULL) {
+    if (!is.null(nbuckets) & is.null(probs)) {
+        breaks <- quantile(x_quantile, probs=0:nbuckets / nbuckets)
+        cut(x_bucket, breaks, include.lowest=TRUE)
+    } else if (!is.null(probs) & is.null(nbuckets)) {
+        cut(x_bucket, quantile(x_quantile, probs), include.lowest=TRUE)
+    } else {
+        stop('Only specify nbuckets or probs, not both.')
+    }
+}
+
+
+#' Compare on- and off-Academia citations within years and quantile groups
+#' of journal impact factors.
+#'
+#' @param cites_df A dataframe with citations and impact factors.
+#' @param summarizer (default mean) A function to summarize citations within groups.
+#' @param comparator (default `/`) A function with arguments (on, off), that compares
+#' on and off-Academia citation summaries. The default computes the on/off ratio.
+#' @param ... Extra parameters to `quantileBuckets`
+#' @return A dataframe with statistic by year, impact factor group, and on/off-source.
+compareByImpactFactorBuckets <- function(cites_df, summarizer=mean,
+                                         comparator=`/`, ...) {
+
+   # Find buckets based on distribution of on-Academia citations.
+   bucketFactors <- function(x) {
+       on_factors <- cites_df %>% filter(source=='on') %>% use_series(impact_factor)
+       quantileBuckets(on_factors, x, ...)
+   }
+
+   cites_df %>%
+        mutate(if_bucket = bucketFactors(impact_factor)) %>%
+        filter(!is.na(if_bucket)) %>%
+        group_by(if_bucket, year, source) %>%
+        summarize(cites=summarizer(citations)) %>%
+        reshape2::dcast(year + if_bucket ~ source, value.var='cites') %>%
+        mutate(comparison = comparator(on, off))
+}
+
+#' Average results over buckets, weighting by the number of on-Academia
+#' articles in the bucket.
+#'
+#' @param cites_df A dataframe of citations with impact factors.
+#' @return A dataframe of weighted average results by year.
+#'
+summarizeOverBuckets <- function(cites_df) {
+    cite_ratios <- compareByImpactFactorBuckets(cites_df,
+                                                summarizer=mean,
+                                                comparator=`/`)
+    counts <- compareByImpactFactorBuckets(cites_df,
+                                           summarizer=length,
+                                           comparator=`+`)
+
+    weights <- counts %>%
+        group_by(year) %>%
+        mutate(weight = on / sum(on)) %>%
+        ungroup %>%
+        select(year, if_bucket, weight)
+
+    cite_ratios %>% left_join(., weights, by=c('year', 'if_bucket')) %>%
+        group_by(year) %>% summarize(wtd_avg = sum(weight * comparison))
+
+}
+
+
+#' Boxplots on- and off-Academia citations within years and quantile groups
+#' of journal impact factors.
+#'
+#' @param cites_df A dataframe with citations and impact factors.
+#' @param ... Extra parameters to `quantileBuckets`
+#'
+#' @return A ggplot2 plot.
+plotByImpactFactorBuckets <- function(cites_df, ...) {
+
+    # Find buckets based on distribution of on-Academia citations.
+    bucketFactors <- function(x) {
+        on_factors <- cites_df %>% filter(source=='on') %>% use_series(impact_factor)
+        quantileBuckets(on_factors, x, ...)
+    }
+
+    # Add bucket variable to data
+    df <- cites_df %>%
+        mutate(if_bucket = bucketFactors(impact_factor)) %>%
+        filter(!is.na(if_bucket))
+
+    p <- ggplot(df, aes(x=factor(year), y=citations, color=source)) +
+        geom_boxplot() +
+        facet_wrap(~if_bucket, ncol=2) +
+        labs(x='Year', y='Citations (log scale)',
+            title='Citations of On- and Off-Academia Articles By Year and Journal Impact Factor') +
+        theme_bw()
+    plotLogScale(p, xy='y')
+}

acadcites/R/figures.R

@@ -0,0 +1,61 @@
+# Reproduce figures
+#
+# Figures:
+# --------
+# 1. Histograms over citations counts by off-/on-Academia
+# 2. Citations boxplots by impact factor bucket and year of publication
+# 3. Scatterplot of cites against impact factor
+# 4. Scatterplot of cites against impact factor by off-/on-Academia and year.
+
+## Function names that produce figures, listed in order
+## of their appearance in the paper.
+.figures_functions = list(
+    'plotCiteDistributions',
+    'plotByImpactFactorBuckets',
+    'plotCitesImpactFactorScatter',
+    'plotImpactFactorMedReg')
+
+
+#' Function to generate figures from the paper.
+#'
+#' Recreate a figure with a given citations dataset by specifying the table's
+#' caption number in the paper.
+#'
+#' @param n Figure caption number.
+#' @param cites_df A data frame with article citations and journal data, as produced by `importData`.
+#' @param ... Optional arguments passed to figure functions.
+#'
+#' @return Nothing. Renders a plot.
+#'
+makeFigure <- function(n, cites_df, ...) {
+    eval(parse(text=.figures_functions[[n]]))(cites_df, ...)
+}
+
+
+plotCiteDistributions <- function(cites_df) {
+    ggplot(cites_df, aes(x=citations)) +
+    geom_histogram(binwidth=1, fill='steelblue', color='white') +
+    xlim(0, 100) +
+    facet_wrap(~source, scales='free_y') +
+    theme_bw()
+}
+
+plotCitesImpactFactorScatter <- function(cites_df) {
+    p <- ggplot(cites_df, aes(x=impact_factor, y=citations)) +
+        geom_point(position=position_jitter(height=.1, width=.01), alpha=.3, size=.75) +
+        geom_smooth(method='lm') +
+        theme_bw() +
+        labs(x='Impact Factor (log scale)', y='Citations (log scale)')
+    plotLogScale(p, c('x', 'y'))
+}
+
+plotImpactFactorMedReg <- function(cites_df) {
+    p <- ggplot(cites_df, aes(x=impact_factor, y=citations, color=source)) +
+        geom_point(position=position_jitter(height=.1, width=.01), alpha=.3, size=.75) +
+        facet_wrap(~year, ncol=2) +
+        stat_quantile(quantiles=0.5) +
+        scale_colour_manual(values=c('orange', 'purple')) +
+        labs(x='Impact Factor (log scale)', y='Citations (log scale)') +
+        theme_bw()
+    plotLogScale(p, c('x', 'y'))
+}