r - Adding a scatterplot to a coefficient plot in ggplot -
i trying create coefficient plot scatterplot of data points laid overtop of it. portion of data listed below:
print(dat) lower upper mode condition obsmean p100 p102 p103 p104 p105 ob.h.black.650 0.693595 0.950305 0.8849853 h.black.650 0.8122302 0.85 1.00 0.85 0.85 0.85 ob.h.white.650 0.649625 0.930375 0.8306119 h.white.650 0.7852518 0.70 0.95 0.80 0.80 0.90 ob.h.black.800 0.833255 0.993105 0.9701335 h.black.800 0.9035971 0.95 1.00 0.95 0.95 1.00 ob.h.white.800 0.802110 0.983230 0.9469604 h.white.800 0.8762590 0.95 0.95 0.80 1.00 0.85 ob.h.black.950 0.869375 0.995800 0.9804180 h.black.950 0.9165468 0.90 1.00 0.85 1.00 1.00 ob.h.white.950 0.862895 0.991750 0.9694811 h.white.950 0.9046763 1.00 1.00 0.95 1.00 0.90 ob.h.black.1100 0.879340 0.996550 0.9885382 h.black.1100 0.9323741 1.00 1.00 0.90 0.75 0.85 ob.h.white.1100 0.853110 0.994405 0.9773240 h.white.1100 0.9140288 0.95 0.95 1.00 0.90 0.90 my code produces output shown below:
ggplot(dat, aes(x = reorder(condition, 1:8), y = mode)) + #x cond, y mode geom_point(size = 4) + #add in black dots modes geom_pointrange(aes(ymin = lower, ymax = upper)) + #add in 95% hdi bars geom_point(aes(x = reorder(condition, 1:8), y = obsmean), size = 4, colour = "red") + geom_point(aes(x = reorder(condition, 1:8), y = p100), #x cond, y ps mean position = position_jitter(width = .2, height = 0), #jitter data point size = 2, colour = "blue") + #make small , blue geom_point(aes(x = reorder(condition, 1:8), y = p102), #x cond, y ps mean position = position_jitter(width = .2, height = 0), #jitter data point size = 2, colour = "blue") + #make small , blue geom_point(aes(x = reorder(condition, 1:8), y = p103), #x condition, y ps mean position = position_jitter(width = .2, height = 0), #jitter data point size = 2, colour = "blue") + #make small , blue geom_point(aes(x = reorder(condition, 1:8), y = p104), #x condition, y ps mean position = position_jitter(width = .2, height = 0), #jitter data point size = 2, colour = "blue") + #make small , blue geom_point(aes(x = reorder(condition, 1:8), y = p105), #x condition, y ps mean position = position_jitter(width = .2, height = 0), #jitter data point size = 2, colour = "blue") + #make small , blue coord_cartesian(ylim = c(.6, 1)) + #scale graph .6 1 ylab("hit rate") + #add y axis label theme_bw() + #use black , white theme theme(axis.title.x = element_blank(), #get rid of x-axis label panel.background = element_blank(), #get rid of background panel.grid = element_blank(), #get rid of gridlines panel.border = element_rect(colour = "black", fill = na, size = 1)) #add border which produces plot.
as can see, code inefficient. each set of points want overlay on condition mean , confidence interval, have enter line of code. 139 repetitions in full dataset, that's particularly cumbersome. wondering if there more efficient way of overlaying individual repetitions on mean , cis.
as mentioned in comments, should reshape data long form. can done either reshape2, data.table or tidyr:
library(reshape2) dat2 <- melt(dat, measure.vars = c("p100","p102","p103","p104","p105")) library(data.table) # method 1 (equivalent reshape2 method) dat2 <- melt(setdt(dat), measure.vars = c("p100","p102","p103","p104","p105")) # method 2 dat2 <- melt(setdt(dat), measure.vars = patterns("^p")) library(tidyr) dat2 <- dat %>% gather(variable, value, c(p100,p102,p103,p104,p105)) next, set factorlevels of condition in correct order:
dat2$condition <- factor(dat2$condition, levels = c("h.black.650","h.white.650","h.black.800","h.white.800", "h.black.950","h.white.950","h.black.1100","h.white.1100")) finally, can simplify plotting code follows:
ggplot(dat2, aes(x = condition, y = mode)) + geom_point(size = 4) + geom_pointrange(aes(ymin = lower, ymax = upper)) + geom_point(aes(x = condition, y = obsmean), size = 4, colour = "red") + geom_point(aes(x = condition, y = value), position = position_jitter(width = .2, height = 0), size = 2, colour = "blue") + coord_cartesian(ylim = c(.6, 1.01)) + ylab("hit rate") + theme_bw() + theme(axis.title.x = element_blank(), panel.background = element_blank(), panel.grid = element_blank(), panel.border = element_rect(colour = "black", fill = na, size = 1)) which results in following plot:
furthermore, might consider coloring blue points p-variables:
ggplot(dat2, aes(x = condition, y = mode)) + geom_point(size = 4) + geom_pointrange(aes(ymin = lower, ymax = upper)) + geom_point(aes(x = condition, y = obsmean), size = 4, colour = "red") + geom_point(aes(x = condition, y = value, color = variable), position = position_jitter(width = .2, height = 0), size = 2) + coord_cartesian(ylim = c(.6, 1.01)) + ylab("hit rate") + theme_bw() + theme(axis.title.x = element_blank(), panel.background = element_blank(), panel.grid = element_blank(), panel.border = element_rect(colour = "black", fill = na, size = 1)) this results in:
Comments
Post a Comment