##' Function to create a related t test plot
##'
##' Function to create a related t test plot
##'
##' Function to create a related t test plot
##' @param predictors the character of the predictor variable
##' @param outcome the character of the outcome variable
##' @param data the dataset
##' @param raw.data should raw data be displayed
##' @param spread type of spread to be displayed
##' @param n the number of datapoints to be displayed
##' @author Dustin Fife
##' @export
= (predictors, outcome, , raw.data, ss, spread="quartiles"){
if ((predictors)!=1){
("Currently, the 'related' option is only available when there's a single predictor.")
}
#### extract levels of the predictors
levs = (,predictors])
if ((levs)!=2){
("Sorry, I can only accept two levels of the grouping variable when related=T.")
}
#### create difference scores
g1 = [data,predictors]==levs[1], outcome]
g2 = [data,predictors]==levs[2], outcome]
if ((g1) != (g2)){
("Sorry, the length of the two groups are not the same. I can only create difference scores when the group sizes are identical.")
}
lab = ("Difference (",levs[2], "-", levs[1], ')')
d2 = (Difference=g2-g1)
if (spread == "stdev"){
summary1 = stat_summary(fun.y='mean', geom='point', size=3, position=position_dodge(=.2))
summary2 = stat_summary(geom='errorbar', fun.ymin = (z){(z)-(z)},
fun.ymax = (z) {(z)+(z)}, fun.y=, size = 1.25, =.12, position=position_dodge(=.2))}
if (spread == "sterr"){
summary1 = stat_summary(fun.y='mean', geom='point', size=3, position=position_dodge(=.2))
summary2 = stat_summary(geom='errorbar', fun.data = mean_cl_normal, color=(1,0,0,.25),
=.12, size = 1.25, position=position_dodge(=.2))}
if (spread == "quartiles"){
summary1 = stat_summary(fun.y='median', geom='point', size=3, position=position_dodge(=.2))
summary2 = stat_summary(geom='errorbar', fun.ymin = (z){(z, .25)},size = 1.25,
fun.ymax = (z) {(z, .75)}, fun.y=, =.12, position=position_dodge(=.2))}
if (ss!=){
m = (d2[sample(1:(d2), size= ss),])
(m) = (d2)
} {
m = d2
}
p = ggplot(d2, aes(y=Difference, x=1)) +
geom_jitter(=m, alpha=(raw.data, .15), =.05) +
summary1 + summary2 +
geom_hline(yintercept=0, ="lightgray") +
labs(x=lab) +
theme_bw() +
theme(axis.ticks.x=element_blank(), axis.text.x=element_blank()) +
coord_cartesian(=(.75, 1.25))
p
}
##' Sample the data
##'
##' Sample the data
##'
##' Sample the data
##' @param ss sample size of raw data to be displayed. Defaults to Inf (for all data)
##' @param data the dataset
##' @author Dustin Fife
##' @export
= (ss, ){
if (ss!=){
m = ([sample(1:(), size= ss),])
(m) = ()
} {
m =
}
}
##' Find transparency of data
##'
##' Find transparency of data
##'
##' Find transparency of data
##' @param raw.data Should raw data be displayed?
##' @param alpha the transparency level
##' @author Dustin Fife
##' @export
= (raw.data,alpha=1){
if (raw.data){
alpha.raw = alpha
} {
alpha.raw = 0
}
alpha.raw
}
##' How to display data
##'
##' How to display data
##'
##' How to display data
##' @param jitter Should data be jittered?
##' @param data the dataset
##' @param ss The size of the sample
##' @param raw.data should raw data be displayed
##' @param alpha the transparency level
##' @author Dustin Fife
##' @export
= (, , ss, raw.data, alpha){
(ss)
if (!()){
if ([1]==T){
jit = geom_jitter(= (ss, ), alpha=(raw.data, alpha=alpha), =.2, =.2)
} if ([1] == F){
jit = geom_point(= (ss, ), alpha=(raw.data, alpha=alpha))
} {
jit = geom_jitter(= (ss, ), alpha=(raw.data, alpha=alpha), =[1], =[2])
}
} {
jit = geom_point(= (ss, ), alpha=(raw.data, alpha=alpha))
}
jit
}
##' Type of spread to use
##'
##' Type of spread to use
##'
##' Type of spread to use
##' @param spread Should be stdev, sterr, or quartiles
##' @param predictors the character of the predictor variable
##' @param outcome the character of the outcome variable
##' @param data the dataset
##' @author Dustin Fife
##' @export
= (spread, predictors, outcome, ){
spread = (spread, ('quartiles', 'stdev', 'sterr'))
if (spread=="stdev"){
summary1 = stat_summary(fun.y='mean', geom='point', size=3, position=position_dodge(=.2))
summary2 = stat_summary(aes_string(x=predictors[1], y=outcome),
geom='errorbar', fun.ymin = (z){(z)-(z)},
fun.ymax = (z) {(z)+(z)}, fun.y=, size = 1.25, =.2, position=position_dodge(=.2))
sum.line = stat_summary(aes_string(=predictors[2]), geom="line", fun.y="mean", position=position_dodge(=.2))
} if (spread=="sterr"){
summary1 = stat_summary(fun.y='mean', geom='point', size=3, position=position_dodge(=.2))
summary2 = stat_summary(aes_string(x=predictors[1], y=outcome), geom='errorbar', fun.data = mean_cl_normal, =.2, size = 1.25, position=position_dodge(=.2))
sum.line = stat_summary(aes_string(=predictors[2]), geom="line", fun.y="mean", position=position_dodge(=.2))
} if (spread == "quartiles"){
summary1 = stat_summary(fun.y='median', geom='point', size=3, position=position_dodge(=.2))
summary2 = stat_summary(aes_string(x=predictors[1], y=outcome), geom='errorbar', fun.ymin = (z){(z, .25)},size = 1.25, fun.ymax = (z) {(z, .75)}, fun.y=, =.2, position=position_dodge(=.2))
sum.line = stat_summary(aes_string(=predictors[2]), geom="line", fun.y="median", position=position_dodge(=.2))
}
(summary1=summary1, summary2=summary2, sum.line=sum.line)
}
##' Type of x summary to use (line or mean)
##'
##' Type of x summary to use (line or mean)
##'
##' Type of x summary to use (line or mean)
##' @param predictors the character of the predictor variable
##' @param outcome the character of the outcome variable
##' @param data the dataset
##' @param spread type of spread to be displayed
##' @author Dustin Fife
##' @export
= (predictors, outcome, spread, ){
if ((,predictors[1]])) {
#### summarize with a line...
#### identify the correct line
if (suppress_smooth){
x_summary = theme_bw()
} if (method=="logistic") {
#### make sure there's only two levels
if (((,outcome]))!=2){
("To fit a logistic curve, you must have only two levels of your outcome variable.")
}
#### convert outcome to numeric (if necessary)
if (!(,outcome])){
,outcome] = (,outcome])-1
}
#### specify the curve
x_summary = geom_smooth(method = "glm", method.args = ( = "binomial"), se = se)
} {
x_summary = geom_smooth(method=method, se=se)
}
} {
##### summarize with dots
x = (spread, predictors, outcome, )
x_summary = (x$summary1,x$summary2)
}
x_summary
}
##' Bin them vars
##'
##' Bin them vars
##'
##' Bin them vars
##' @param given stuff
##' @param bins The number of bins used when putting quantitative variables into categories. A list can be used if different amounts are wanted for different variables
##' @param labels A list of the same length as bins
##' @param breaks The breaks to be used for the bins
##' @param data the dataset
##' @param ghost.reference What should the reference group be (from which to draw the ghost line)? The user can specify up to two values as a vector. See examples.
##' @param prediction The predicted line
##' @author Dustin Fife
##' @export
= (given, breaks, , , bins, ghost.reference, prediction){
##### only allow two "given" variables
if ((given)>2){
("Only two 'given' variables are allowed.")
}
#### modify given (if needed)
if ((given)>0 & !(given)){
if (("+", given)){
given = ((given, " + ", fixed=T))
}
}
#### see if more than two variables are shown at the left of the given sign
if (((predictors) - (given))>2){
("Only two 'axis' variables are allowed.")
}
#### make given variables ggplot friendly (for facet_grid)
given.as.string = ((given)>1 & !(given),((given), collapse="~"), ("~",given))
### identify the number of binned variables we need
if (()>1 & [2] %in% numbers){
binned.vars = ([2], numbers[which((numbers) %in% given)])
} {
binned.vars = numbers[which((numbers) %in% given)]
}
if ((binned.vars)>0){
msg = ("The following variables are going to be binned: ", (binned.vars, collapse=", "), "\n")
(msg)
}
### repeat the bins the number of bins there are
if ((bins) != (binned.vars) & (bins)>1){
("You haven't specified enough bins to cover all the binned variables. I'm making a guess for the rest of the variables")
bins = (bins, =1, =(binned.vars))
}
if ((bins)==1){
bins = (bins, times=(binned.vars))
}
#### bin the binned variables
if ((binned.vars)>0){
for (i 1:(binned.vars)){
break.current = (breaks[i])
if (!(([i])) & (([i])) != bins[i]){
(("The label vectors (", (([i]), collapse=", "), ") is not the same length as the bin length (", bins[i], ")", sep=""))
}
### if they supply the breaks...
if (!(break.current)){
#### give min as breaks, if the user doesn't
if ((break.current)>(,binned.vars[i]])){
break.current = (-, break.current)
}
if ((break.current,na.rm=T)<(,binned.vars[i]])){
break.current = (break.current, )
}
quants = (break.current)
} {
quants = (,binned.vars[i]], (from=0, to=1, length.out=bins[i]+1))
}
,(binned.vars[i])] = (,binned.vars[i]], quants, = ([i]), include.lowest=T, include.highest=T)
#### if they're making a ghost reference, bin that too
if (!(ghost.reference) & binned.vars[i] %in% (ghost.reference)){
val = (ghost.reference[binned.vars[i]])
ghost.reference[binned.vars[i]] = ((val, quants, =([i]), include.lowest=T, include.highest=T))
}
if (!(prediction)){
prediction,(binned.vars[i])] = (prediction,binned.vars[i]], quants, = ([i]), include.lowest=T, include.highest=T)
}
}
if (!(prediction)){
#### average the predictions within bin
f = ("prediction", ("model",
predictors-(predictors==binned.vars[i])],
binned.vars[i]
)
)
prediction = (f, =prediction, FUN=)
}
#### add code for "given" variable
}
if (!(given[1])){
giv = facet_grid((given.as.string),labeller = labeller(.rows = label_both, .cols=label_both))
} {
giv = theme_bw()
}
(prediction=prediction, ghost.reference=ghost.reference, =, giv=giv)
}
