Please be tolerant :) I am a dummy user of R and I am using the code and sample data to learn how to make forest plot that was shown in the previous post -
Optimal/efficient plotting of survival/regression analysis results
I was wondering is it possible to set user-defined x-axis scale with the code shown there? Up to now x a-axis scale is defined somehow automatically.
Thank you for any tips.
I'm unimpressed with the precision of the documentation since one might assume that the limits argument would be values on the relative risk scale rather than on the log-transformed scale. One gets a ridiculous result if that is done. That quibble not withstanding, it's relatively easy to use that parameter to created an expanded plot:
install('devtools') # then use it to get current package
# executing the install and load of the package referenced at the top of that answer
print(forest_model(lung_cox, limits=log( c(.5, 50) ) ))
Trying for a lower range of 0 on the relative risk scale is not sensible. Would imply a -Inf value on hte log-transformed scale. Trying for lower value, say log(0.001), confuses the pretty printing of the scale in my tests.
Related
everyone!
I need help with one function in R (cv.tree). I created a regression tree, and now I need to find the optimal alpha parameter and its corresponding subtree for pruning. I know, that usually we use cross-validation, but I don't understand well what is shown on the graph on the y-axis. It says that on the x-axis is the size of each tree in the sequence. On the y-axis, cv.tree$dev is shown. And I don't understand what it is. I will be grateful for help. (Maybe you can also explain, how to calculate this value).
I created a logistic/logit model with a binomial response variable using
model <- glm(response~predictor1+predictor2+...)
and then I used the predict function to create a new data frame
outcome <-data.frame(predict(model,newdata=IndependentDataSet,type="response"),as.numeric(as.character(Independent$ResponseVariable)))
names(outcome) <- c("Pr","Obs")
I can use one of the following functions
plot(verify(data$obs,data$pr),CI=TRUE)
attribute(verify(data$obs,data$pr))
to create a plot that looks like this
or
reliability.plot(verify(data$obs,data$pr))
from
library(verification)
to create a reliability diagram. I am wondering how I can separate the bins based on specific values. For example, the model that I am evaluating is based around a climatology of 19% (0.19) and I want there to be a bin at (1/3)*climatology, climatology, and go up by (2/3) of climatology for the proceeding bins. How can I do this?
Additionally, I have seen the bins represented as circles that are proportional in size to the percent of the data that is at that bin. Does anyone know how to make a more aesthetically pleasing reliability diagram in R? Any recommendations are welcome.
This is how I would like my diagrams to appear
The easiest could be using
trace("attribute.default",edit=TRUE)
or whichever other function.
In this way, you access the source code and edit it. These changes affect only the current R session.
I started to really like violin plots, since they give me a much better feel that box plots when you have funny distributions. I like to automatize a lot of stuff, and thus ran into a problem:
When one variable has 0 variance, the boxplot just gives you a line at that point. Geom_violin however, terminates with an error. What behavior would I like? Well, either put in a line or nothing, but please give me the distributions for the other variables.
Ok, quick example:
dff=data.frame(x=factor(rep(1:2,each=100)),y=c(rnorm(100),rep(0,100)))
ggplot(dff,aes(x=x,y=y)) + geom_violin()
yields
Error in `$<-.data.frame`(`*tmp*`, "n", value = 100L) :
replacement has 1 row, data has 0
However, what works is:
ggplot(dff,aes(x=x,y=y)) + geom_boxplot()
Update:
The issue is resolved as of yesterday: https://github.com/hadley/ggplot2/issues/972
Update 2:
(from question author)
Wow, Hadley himself responded! geom_violin now behaves consistently with geom_density and base R density.
However, I don't think the behavior is optimal yet.
(1) The 'zero' problem
Just run it with my original example:
dff=data.frame(x=factor(rep(1:2, each=100)), y=c(rnorm(100), rep(0,100)))
ggplot(dff,aes(x=x,y=y)) + geom_violin(trim=FALSE)
Yielding this:
Is the plot on the right an appropriate representation of 'all zeroes'? I don't think so. It is better to have trimming that produces a single line to show that there is no variation in the data.
Workaround solution: Add a + geom_boxplot()
(2) I may actually want TRIM=TRUE.
Example:
dff=data.frame(x=factor(rep(1:2, each=100)), y=c(rgamma(100,1,1), rep(0,100) ))
ggplot(dff,aes(x=x,y=y)) + geom_violin(trim=FALSE)
Now I have non-zero data, and standard kernel density estimates don't handle this correctly. With trim=T I can quickly see that the data is strictly positive.
I am not arguing that the current behavior is 'wrong', since it's in line with other functions. However, geom_violin may be used in different contexts, for exploring different data.frames with heterogeneous data types (positive+skewed or not, for instance).
Three options for dealing with this until the ggplot2 issue is resolved:
As a quick hack, you can set one of the y-values to 0.0001 (instead of zero) and geom_violin will work.
Check out the vioplot package if you're not set on using ggplot2. vioplot doesn't throw an error when you feed it a bunch of identical values.
The Hmisc package includes a panel.bpplot (box-percentile plot) function that can create violin plots with the bwplot function from the lattice package. See the Examples section of ?panel.bpplot. It produces a single line when you feed it a vector of identical values.
I'm trying to create an actuarial survival analysis in R (I'm following some worked examples). I think the best way to do this is using the survival package. So something like:
library(survival)
surv.test <- survfit(Surv(TIME,STATUS), data=test)
However, to get the correct answer I will need to divide the TIME variable into 365 day intervals and I can't quite work out how to do this so it matches the given result.
As far as I can make out, there is no option within the survfit function that will do this. I went through several document examples and none of them were trying to create a stairstep type of plot (there is a type='interval' option, but seems to do something different). So I guess I need to regroup my data before I apply the survival function?
Any ideas?
P.S: In SPSS this would be INTERVAL = THRU 10000 BY 365; in Stata intervals(365) ... connect(stairsteps)
I am guessing that you want to divide the TIME variable into intervals because you want to plot a Kaplan-Meier curve. In R, that isn't necessary, you can just call plot on the survfit object. For example,
s=survfit(Surv(futime, fustat)~rx, data=ovarian)
plot(s)
I think I understand your question a little better. The reason why you are getting a thick black line is because you have a lot of censoring, and a + is being plotted at every single point where there is censoring, you can turn this off with mark.time=F. (You can see other options in ?survival:::plot.survfit)
However, if you still want to aggregate by year, simply divide your follow up time by 365, and round up. ceiling is used to round up. Here is an example of aggregating at different time levels without censoring.
par(mfrow=c(1,3))
plot(survfit(Surv(ceiling(futime), fustat)~rx, data=ovarian),col=c('blue','red'),main='Day',mark.time=F)
plot(survfit(Surv(ceiling(futime/30), fustat)~rx, data=ovarian),col=c('blue','red'),main='Month',mark.time=F)
plot(survfit(Surv(ceiling(futime/365), fustat)~rx, data=ovarian),col=c('blue','red'),main='Year',mark.time=F)
par(mfrow=c(1,1))
But I think that plotting the Kaplan-Meier without the censoring symbols will look very nice, and provide more insight.
Hurray, I should be able to post the images now:
1) this is how the R basic survival plot looks like at the moment
2) and this is how it should look like (SPSS example)
That was exactly what I was missing! Thanks!
Solution:
vas.surv <- survfit(Surv(ceiling(TIME/365), STATUS)~1, conf.type="none", data=vasectomy)
plot(vas.surv, ylim=c(0.975,1), mark.time=F, xlab="Years", ylab="Cumulative Survival")
A nice touch would be to displays the days on the x-axis instead of the years (as in SPSS) example, but I'm not too bothered about this.
Background:
I'm running a Monte Carlo simulation to show that a particular process (a cumulative mean) does not converge over time, and often diverges wildly in simulation (the expectation of the random variable = infinity). I want to plot about 10 of these simulations on a line chart, where the x axis has the iteration number, and the y axis has the cumulative mean up to that point.
Here's my problem:
I'll run the first simulation (each sim. having 10,000 iterations), and build the main plot based on its current range. But often one of the simulations will have a range a few orders of magnitude large than the first one, so the plot flies outside of the original range. So, is there any way to dynamically update the ylim or xlim of a plot upon adding a new set of points or lines?
I can think of two workarounds for this: 1. store each simulation, then pick the one with the largest range, and build the base graph off of that (not elegant, and I'd have to store a lot of data in memory, but would probably be laptop-friendly [[EDIT: as Marek points out, this is not a memory-intense example, but if you know of a nice solution that'd support far more iterations such that it becomes an issue (think high dimensional walks that require much, much larger MC samples for convergence) then jump right in]]) 2. find a seed that appears to build a nice looking version of it, and set the ylim manually, which would make the demonstration reproducible.
Naturally I'm holding out for something more elegant than my workarounds. Hoping this isn't too pedestrian a problem, since I imagine it's not uncommon with simulations in R. Any ideas?
I'm not sure if this is possible using base graphics, if someone has a solution I'd love to see it. However graphics systems based on grid (lattice and ggplot2) allow the graphics object to be saved and updated. It's insanely easy in ggplot2.
require(ggplot2)
make some data and get the range:
foo <- as.data.frame(cbind(data=rnorm(100), numb=seq_len(100)))
make an initial ggplot object and plot it:
p <- ggplot(as.data.frame(foo), aes(numb, data)) + layer(geom='line')
p
make some more data and add it to the plot
foo <- as.data.frame(cbind(data=rnorm(200), numb=seq_len(200)))
p <- p + geom_line(aes(numb, data, colour="red"), data=as.data.frame(foo))
plot the new object
p
I think (1) is the best option. I actually don't think this isn't elegant. I think it would be more computationally intensive to redraw every time you hit a point greater than xlim or ylim.
Also, I saw in Peter Hoff's book about Bayesian statistics a cool use of ts() instead of lines() for cumulative sums/means. It looks pretty spiffy: