mkiang/google_cites.R

## google_cites.R
## Code for this blog post:
## http://mathewkiang.com/2016/10/08/a-visual-tour-of-my-publications/

## Imports
library(RColorBrewer)
library(ggplot2)
library(dplyr)
library(devtools)
devtools::install_github("jkeirstead/scholar")
library(scholar)

## Helper function
get_all_article_history <- function (id, sleep = 0, return_pubs = FALSE,
                                     debug = FALSE) {
    ## Get publications and unique articles (with citations)
    pubs <- get_publications(id)
    articles <- as.character(unique(pubs$pubid[pubs$cites > 0]))
    npubs <- length(articles)

    ## Loop through each article and get citation history
    holder <- NULL
    i <- 1
    for (pub in articles) {
        if (debug) {
            print(paste0("(", i, " of ", npubs, ") ", pub, " | ", Sys.time()))
        }

        x <- get_article_cite_history(id = id, article = pub)
        holder <- rbind(holder, x)

        if (sleep > 0) {
            ## Google Scholar doesn't have a real API. To avoid violating TOS,
            ## we space out requests by a specified number of minutes (20 - 30)
            ## and add some randomness. (abs() controls for unlikely event of
            ## negative number.)
            sleeptime <- 60 * abs(rnorm(1, sleep, .25 * sleep))
            Sys.sleep(sleeptime)
        }

        i <- i + 1
    }

    if (sum(pubs$cites == 0) > 0) {
        holder <- rbind(holder, pubs[pubs$cites == 0,
                                     c("year", "cites", "pubid")])
    }

    ## Return both the author history and the citation history
    if (return_pubs) {
        return(list(pubs = pubs, cites = holder))
    }
    return(holder)
}

## Define author
mvk <- "eD9_J3wAAAAJ"

## Get publication and citation history
mk <- get_all_article_history(mvk, return_pubs = TRUE)

## Extract publications and then merge with the type of paper
## Note: All publications as of 10/6/2016
mkpubs <- mk$pubs
pubid <- c("d1gkVwhDpl0C", "u-x6o8ySG0sC", "qjMakFHDy7sC", "9yKSN-GCB0IC",
           "YsMSGLbcyi4C", "Y0pCki6q_DkC", "zYLM7Y9cAGgC", "u5HHmVD_uO8C",
           "Tyk-4Ss8FVUC", "eQOLeE2rZwMC", "IjCSPb-OGe4C", "2osOgNQ5qMEC",
           "L_l9e5I586QC", "hqOjcs7Dif8C", "LkGwnXOMwfcC", "WF5omc3nYNoC",
           "__bU50VfleQC", "roLk4NBRz8UC", "_FxGoFyzp5QC", "D_tqNUsBuKoC",
           "tHtfpZlB6tUC", "c1e4I3QdEKYC")
ptypes <- c("MSM/Drugs", "MSM/Drugs", "MSM/Drugs", "HPM",
           "Inequality", "Inequality", "HPM", "MSM/Drugs",
           "MSM/Drugs", "Inequality", "MSM/Drugs", "MSM/Drugs",
           "Inequality", "Inequality", "HPM", "HPM",
           "DP", "Inequality", "Inequality", "Inequality",
           "HPM", "HPM")
pubtypes <- data.frame(pubid = pubid, pubtype = ptypes)

## Clean up factors with more descriptive names
pubtypes$pubtype <- factor(pubtypes$pubtype,
                           levels = c("DP", "Inequality",
                                      "HPM", "MSM/Drugs"),
                           labels = c("Digital Phenotyping",
                                      "Health Inequalities",
                                      "Health Policy and Management",
                                      "MSM / HIV / Drugs"),
                           ordered = TRUE)
mkpubs <- merge(mkpubs, pubtypes, by = "pubid")

## Rename column for merging, then merge, and get years from publication
names(mkpubs)[grep(x = names(mkpubs), pattern = "year")] <- "pubyear"
mkcites <- mk$cites
mkcites <- merge(mkcites, select(mkpubs, pubid, pubyear, pubtype),
                 by = "pubid")
mkcites$delta <- ifelse(mkcites$year - mkcites$pubyear >= 0,
                        mkcites$year - mkcites$pubyear, 0)

## Get cumulative sum -- remove NAs
mkcites %<>%
    group_by(pubid) %>%
    arrange(year) %>%
    mutate(ccite = cumsum(cites))
mkcites <- mkcites[!is.na(mkcites$pubyear), ]

## If delta doesn't start at zero, add it in.
for (id in unique(mkcites$pubid)) {
    if (min(mkcites$delta[mkcites$pubid == id], na.rm = TRUE) != 0) {
        pyear <- mkcites$pubyear[mkcites$pubid == id][1]
        ptype <- mkcites$pubtype[mkcites$pubid == id][1]
        ## Can't just use c() below since pubid is a string
        new_row <- data.frame(pubid = id, year = pyear, cites = 0,
                              pubyear = pyear, pubtype = ptype,
                                  delta = 0, ccite = 0)
        ## tibbles handle rbinding strangely -- convert back and forth
        mkcites <- as_data_frame(rbind(as.data.frame(mkcites), new_row))
    }
}


## Feature image
p0 <- ggplot(data = mkcites, aes(x = year, y = ccite)) +
    geom_line(stat = "smooth", method = "lm", se = FALSE,
              color = "gray60", size = 1.5, alpha = .15) +
    geom_point(aes(group = pubid),size = 1, alpha = .75, color = "grey30") +
    geom_line(aes(group = pubid), size = .75, alpha = .75, color = "grey30") +
    theme_classic() +
    scale_x_continuous("", expand = c(0, .07)) +
    scale_y_continuous("Citations", expand = c(0, .15)) +
    labs(title = "Citation history over time")

ggsave(p0, filename = "./feature_img.jpg", width = 7, height = 4, scale = 1.2)

## Plot of each article's citations over age of article
p1 <- ggplot(data = mkcites, aes(x = delta, y = ccite, color = pubtype)) +
    geom_line(stat = "smooth", method = "lm", se = FALSE,
              color = "gray60", size = 1.5, alpha = .5) +
    geom_point(aes(group = pubid),size = 1, alpha = .75) +
    geom_line(aes(group = pubid), size = .75, alpha = .75) +
    theme_classic() +
    scale_color_brewer("Article Type", type = "qual", palette = "Dark2") +
    scale_x_continuous("Years from publication", expand = c(0, .02)) +
    scale_y_continuous("Citations", expand = c(0, .15)) +
    theme(legend.position = c(0, 1),
          legend.justification = c(0, 1)) +
    labs(title = "Citation trajectory of my articles")

ggsave(p1, filename = "./mk_traj_shift.jpg", width = 7, height = 4, scale = 1.2)


## Plot of each article's citation over time
p2 <- ggplot(data = mkcites, aes(x = year, y = ccite, color = pubtype)) +
    geom_line(stat = "smooth", method = "lm", se = FALSE,
              color = "gray60", size = 1.5, alpha = .85) +
    geom_point(aes(group = pubid),size = 1, alpha = .75) +
    geom_line(aes(group = pubid), size = .75, alpha = .75) +
    theme_classic() +
    scale_color_brewer("Article Type", type = "qual", palette = "Dark2") +
    scale_x_continuous("", expand = c(0, .07)) +
    scale_y_continuous("Citations", expand = c(0, .15)) +
    theme(legend.position = c(0, 1),
          legend.justification = c(0, 1)) +
    labs(title = "Citation history over time")

ggsave(p2, filename = "./mk_traj.jpg", width = 7, height = 4, scale = 1.2)


## Same as p2 but aggregate by article type
p3 <- ggplot(data = mkcites, aes(x = year, y = ccite)) +
    geom_point(aes(group = pubid), color = "grey75",size = 1, alpha = .25) +
    geom_line(aes(group = pubid), color = "grey75", size = .75, alpha = .25) +
    geom_line(stat = "smooth", method = "lm", se = FALSE,
              aes(color = pubtype), size = 1.5, alpha = .9) +
    theme_classic() +
    scale_color_brewer("Article Type", type = "qual", palette = "Dark2") +
    scale_x_continuous("", expand = c(0, .07)) +
    scale_y_continuous("Citations", expand = c(0, .15), limits = c(0, 25)) +
    theme(legend.position = c(0, 1),
          legend.justification = c(0, 1)) +
    labs(title = "Citation history, fitted by type, over time")

ggsave(p3, filename = "./mk_traj_lm.jpg", width = 7, height = 4, scale = 1.2)


## Same as p1 but again aggregate by article type
p4 <- ggplot(data = mkcites, aes(x = delta, y = ccite, color = pubtype)) +
    geom_point(aes(group = pubid), color = "grey75", size = 1, alpha = .25) +
    geom_line(aes(group = pubid), color = "grey75", size = .75, alpha = .25) +
    geom_line(stat = "smooth", method = "lm", se = FALSE,
              aes(color = pubtype), size = 1.5, alpha = .9) +
    theme_classic() +
    scale_color_brewer("Article Type", type = "qual", palette = "Dark2") +
    scale_x_continuous("Years from publication", expand = c(0, .02)) +
    scale_y_continuous("Citations", expand = c(0, .15)) +
    theme(legend.position = c(0, 1),
          legend.justification = c(0, 1)) +
    labs(title = "Citation history, fitted by type, over article's age")

ggsave(p4, filename = "./mk_traj_shift_lm.jpg", width = 7, height = 4, scale = 1.2)


## Reshape dataframe for stacked barcharts
mkstack <- mkcites %>%
    group_by(year, pubtype) %>%
    summarize(total = sum(cites))

mkstackc <- mkcites %>%
    group_by(year, pubtype) %>%
    summarize(total = sum(ccite))

## Citations per year by group
p5 <- ggplot(data = mkstack,
             aes(x = year, y = total, fill = pubtype)) +
    geom_bar(stat="identity", alpha = .9) +
    geom_hline(yintercept = seq(0, 65, 10), color = "white", alpha = .75) +
    theme_classic() +
    scale_fill_brewer("Article Type", type = "qual", palette = "Dark2") +
    scale_x_continuous("", expand = c(0, .02),
                       breaks = 2010:2016, labels = 2010:2016) +
    scale_y_continuous("Citations", expand = c(0, .15),
                       breaks = seq(0, 65, 10), labels = seq(0, 65, 10)) +
    theme(legend.position = c(0, 1),
          legend.justification = c(0, 1)) +
    labs(title = "Citations per year by article type")

ggsave(p5, filename = "./mk_bar.jpg", width = 4, height = 4, scale = 1.2)

## Cumulative citations per year by group
p6 <- ggplot(data = mkstackc,
             aes(x = year, y = total, fill = pubtype)) +
    geom_bar(stat="identity", alpha = .9) +
    geom_hline(yintercept = seq(0, 200, 25), color = "white", alpha = .75) +
    theme_classic() +
    scale_fill_brewer("Article Type", type = "qual", palette = "Dark2") +
    scale_x_continuous("", expand = c(0, .02),
                       breaks = 2010:2016, labels = 2010:2016) +
    scale_y_continuous("Cumulative citations", expand = c(0, .15),
                       breaks = seq(0, 200, 25), labels = seq(0, 200, 25)) +
    theme(legend.position = c(0, 1),
          legend.justification = c(0, 1)) +
    labs(title = "Cumulative citations per year by article type")

ggsave(p6, filename = "./mk_bar_cum.jpg", width = 4, height = 4, scale = 1.2)
	## Code for this blog post:
	## http://mathewkiang.com/2016/10/08/a-visual-tour-of-my-publications/

	## Imports
	library(RColorBrewer)
	library(ggplot2)
	library(dplyr)
	library(devtools)
	devtools::install_github("jkeirstead/scholar")
	library(scholar)

	## Helper function
	get_all_article_history <- function (id, sleep = 0, return_pubs = FALSE,
	debug = FALSE) {
	## Get publications and unique articles (with citations)
	pubs <- get_publications(id)
	articles <- as.character(unique(pubs$pubid[pubs$cites > 0]))
	npubs <- length(articles)

	## Loop through each article and get citation history
	holder <- NULL
	i <- 1
	for (pub in articles) {
	if (debug) {
	print(paste0("(", i, " of ", npubs, ") ", pub, " \| ", Sys.time()))
	}

	x <- get_article_cite_history(id = id, article = pub)
	holder <- rbind(holder, x)

	if (sleep > 0) {
	## Google Scholar doesn't have a real API. To avoid violating TOS,
	## we space out requests by a specified number of minutes (20 - 30)
	## and add some randomness. (abs() controls for unlikely event of
	## negative number.)
	sleeptime <- 60 * abs(rnorm(1, sleep, .25 * sleep))
	Sys.sleep(sleeptime)
	}

	i <- i + 1
	}

	if (sum(pubs$cites == 0) > 0) {
	holder <- rbind(holder, pubs[pubs$cites == 0,
	c("year", "cites", "pubid")])
	}

	## Return both the author history and the citation history
	if (return_pubs) {
	return(list(pubs = pubs, cites = holder))
	}
	return(holder)
	}

	## Define author
	mvk <- "eD9_J3wAAAAJ"

	## Get publication and citation history
	mk <- get_all_article_history(mvk, return_pubs = TRUE)

	## Extract publications and then merge with the type of paper
	## Note: All publications as of 10/6/2016
	mkpubs <- mk$pubs
	pubid <- c("d1gkVwhDpl0C", "u-x6o8ySG0sC", "qjMakFHDy7sC", "9yKSN-GCB0IC",
	"YsMSGLbcyi4C", "Y0pCki6q_DkC", "zYLM7Y9cAGgC", "u5HHmVD_uO8C",
	"Tyk-4Ss8FVUC", "eQOLeE2rZwMC", "IjCSPb-OGe4C", "2osOgNQ5qMEC",
	"L_l9e5I586QC", "hqOjcs7Dif8C", "LkGwnXOMwfcC", "WF5omc3nYNoC",
	"__bU50VfleQC", "roLk4NBRz8UC", "_FxGoFyzp5QC", "D_tqNUsBuKoC",
	"tHtfpZlB6tUC", "c1e4I3QdEKYC")
	ptypes <- c("MSM/Drugs", "MSM/Drugs", "MSM/Drugs", "HPM",
	"Inequality", "Inequality", "HPM", "MSM/Drugs",
	"MSM/Drugs", "Inequality", "MSM/Drugs", "MSM/Drugs",
	"Inequality", "Inequality", "HPM", "HPM",
	"DP", "Inequality", "Inequality", "Inequality",
	"HPM", "HPM")
	pubtypes <- data.frame(pubid = pubid, pubtype = ptypes)

	## Clean up factors with more descriptive names
	pubtypes$pubtype <- factor(pubtypes$pubtype,
	levels = c("DP", "Inequality",
	"HPM", "MSM/Drugs"),
	labels = c("Digital Phenotyping",
	"Health Inequalities",
	"Health Policy and Management",
	"MSM / HIV / Drugs"),
	ordered = TRUE)
	mkpubs <- merge(mkpubs, pubtypes, by = "pubid")

	## Rename column for merging, then merge, and get years from publication
	names(mkpubs)[grep(x = names(mkpubs), pattern = "year")] <- "pubyear"
	mkcites <- mk$cites
	mkcites <- merge(mkcites, select(mkpubs, pubid, pubyear, pubtype),
	by = "pubid")
	mkcites$delta <- ifelse(mkcites$year - mkcites$pubyear >= 0,
	mkcites$year - mkcites$pubyear, 0)

	## Get cumulative sum -- remove NAs
	mkcites %<>%
	group_by(pubid) %>%
	arrange(year) %>%
	mutate(ccite = cumsum(cites))
	mkcites <- mkcites[!is.na(mkcites$pubyear), ]

	## If delta doesn't start at zero, add it in.
	for (id in unique(mkcites$pubid)) {
	if (min(mkcites$delta[mkcites$pubid == id], na.rm = TRUE) != 0) {
	pyear <- mkcites$pubyear[mkcites$pubid == id][1]
	ptype <- mkcites$pubtype[mkcites$pubid == id][1]
	## Can't just use c() below since pubid is a string
	new_row <- data.frame(pubid = id, year = pyear, cites = 0,
	pubyear = pyear, pubtype = ptype,
	delta = 0, ccite = 0)
	## tibbles handle rbinding strangely -- convert back and forth
	mkcites <- as_data_frame(rbind(as.data.frame(mkcites), new_row))
	}
	}


	## Feature image
	p0 <- ggplot(data = mkcites, aes(x = year, y = ccite)) +
	geom_line(stat = "smooth", method = "lm", se = FALSE,
	color = "gray60", size = 1.5, alpha = .15) +
	geom_point(aes(group = pubid),size = 1, alpha = .75, color = "grey30") +
	geom_line(aes(group = pubid), size = .75, alpha = .75, color = "grey30") +
	theme_classic() +
	scale_x_continuous("", expand = c(0, .07)) +
	scale_y_continuous("Citations", expand = c(0, .15)) +
	labs(title = "Citation history over time")

	ggsave(p0, filename = "./feature_img.jpg", width = 7, height = 4, scale = 1.2)

	## Plot of each article's citations over age of article
	p1 <- ggplot(data = mkcites, aes(x = delta, y = ccite, color = pubtype)) +
	geom_line(stat = "smooth", method = "lm", se = FALSE,
	color = "gray60", size = 1.5, alpha = .5) +
	geom_point(aes(group = pubid),size = 1, alpha = .75) +
	geom_line(aes(group = pubid), size = .75, alpha = .75) +
	theme_classic() +
	scale_color_brewer("Article Type", type = "qual", palette = "Dark2") +
	scale_x_continuous("Years from publication", expand = c(0, .02)) +
	scale_y_continuous("Citations", expand = c(0, .15)) +
	theme(legend.position = c(0, 1),
	legend.justification = c(0, 1)) +
	labs(title = "Citation trajectory of my articles")

	ggsave(p1, filename = "./mk_traj_shift.jpg", width = 7, height = 4, scale = 1.2)


	## Plot of each article's citation over time
	p2 <- ggplot(data = mkcites, aes(x = year, y = ccite, color = pubtype)) +
	geom_line(stat = "smooth", method = "lm", se = FALSE,
	color = "gray60", size = 1.5, alpha = .85) +
	geom_point(aes(group = pubid),size = 1, alpha = .75) +
	geom_line(aes(group = pubid), size = .75, alpha = .75) +
	theme_classic() +
	scale_color_brewer("Article Type", type = "qual", palette = "Dark2") +
	scale_x_continuous("", expand = c(0, .07)) +
	scale_y_continuous("Citations", expand = c(0, .15)) +
	theme(legend.position = c(0, 1),
	legend.justification = c(0, 1)) +
	labs(title = "Citation history over time")

	ggsave(p2, filename = "./mk_traj.jpg", width = 7, height = 4, scale = 1.2)


	## Same as p2 but aggregate by article type
	p3 <- ggplot(data = mkcites, aes(x = year, y = ccite)) +
	geom_point(aes(group = pubid), color = "grey75",size = 1, alpha = .25) +
	geom_line(aes(group = pubid), color = "grey75", size = .75, alpha = .25) +
	geom_line(stat = "smooth", method = "lm", se = FALSE,
	aes(color = pubtype), size = 1.5, alpha = .9) +
	theme_classic() +
	scale_color_brewer("Article Type", type = "qual", palette = "Dark2") +
	scale_x_continuous("", expand = c(0, .07)) +
	scale_y_continuous("Citations", expand = c(0, .15), limits = c(0, 25)) +
	theme(legend.position = c(0, 1),
	legend.justification = c(0, 1)) +
	labs(title = "Citation history, fitted by type, over time")

	ggsave(p3, filename = "./mk_traj_lm.jpg", width = 7, height = 4, scale = 1.2)


	## Same as p1 but again aggregate by article type
	p4 <- ggplot(data = mkcites, aes(x = delta, y = ccite, color = pubtype)) +
	geom_point(aes(group = pubid), color = "grey75", size = 1, alpha = .25) +
	geom_line(aes(group = pubid), color = "grey75", size = .75, alpha = .25) +
	geom_line(stat = "smooth", method = "lm", se = FALSE,
	aes(color = pubtype), size = 1.5, alpha = .9) +
	theme_classic() +
	scale_color_brewer("Article Type", type = "qual", palette = "Dark2") +
	scale_x_continuous("Years from publication", expand = c(0, .02)) +
	scale_y_continuous("Citations", expand = c(0, .15)) +
	theme(legend.position = c(0, 1),
	legend.justification = c(0, 1)) +
	labs(title = "Citation history, fitted by type, over article's age")

	ggsave(p4, filename = "./mk_traj_shift_lm.jpg", width = 7, height = 4, scale = 1.2)


	## Reshape dataframe for stacked barcharts
	mkstack <- mkcites %>%
	group_by(year, pubtype) %>%
	summarize(total = sum(cites))

	mkstackc <- mkcites %>%
	group_by(year, pubtype) %>%
	summarize(total = sum(ccite))

	## Citations per year by group
	p5 <- ggplot(data = mkstack,
	aes(x = year, y = total, fill = pubtype)) +
	geom_bar(stat="identity", alpha = .9) +
	geom_hline(yintercept = seq(0, 65, 10), color = "white", alpha = .75) +
	theme_classic() +
	scale_fill_brewer("Article Type", type = "qual", palette = "Dark2") +
	scale_x_continuous("", expand = c(0, .02),
	breaks = 2010:2016, labels = 2010:2016) +
	scale_y_continuous("Citations", expand = c(0, .15),
	breaks = seq(0, 65, 10), labels = seq(0, 65, 10)) +
	theme(legend.position = c(0, 1),
	legend.justification = c(0, 1)) +
	labs(title = "Citations per year by article type")

	ggsave(p5, filename = "./mk_bar.jpg", width = 4, height = 4, scale = 1.2)

	## Cumulative citations per year by group
	p6 <- ggplot(data = mkstackc,
	aes(x = year, y = total, fill = pubtype)) +
	geom_bar(stat="identity", alpha = .9) +
	geom_hline(yintercept = seq(0, 200, 25), color = "white", alpha = .75) +
	theme_classic() +
	scale_fill_brewer("Article Type", type = "qual", palette = "Dark2") +
	scale_x_continuous("", expand = c(0, .02),
	breaks = 2010:2016, labels = 2010:2016) +
	scale_y_continuous("Cumulative citations", expand = c(0, .15),
	breaks = seq(0, 200, 25), labels = seq(0, 200, 25)) +
	theme(legend.position = c(0, 1),
	legend.justification = c(0, 1)) +
	labs(title = "Cumulative citations per year by article type")

	ggsave(p6, filename = "./mk_bar_cum.jpg", width = 4, height = 4, scale = 1.2)