What is a good alternative to boxplots that reflects the MEAN and STANDARD ERROR , not the median and its quartiles ? A package in R that does that would help a lot.
3 Answers
You could just use a barchart with error bars that represent the standard error. Depending on your data structure you may have to summarise it and calculate the standard errors using plyr
I quite like the ggplot geom_pointrange for non-standard point and range type plots. You supply the point and the min and max values yourself, so as long as you explicit about what you are using the range to represent (std. error, variance, quantiles, etc), you can use the geom to plot whatever you need. Purists might disagree, and argue that the point_range geom should only be used to plot the complete range of the data, if so I'd be interested to know this.
One of the advantages of R is that you can a) Modify existing code and b) Make your own graphs "from scratch" using either base graphics or one of the packages that have been developed.
Questions about code are off topic here, but it would surely not be too difficult to either modify boxplot
or build your own function.
I'd be hesitant to make it look too much like a boxplot, however, because, even if you label things clearly as mean and standard error, people are so used to boxplots that they will interpret it as one.