These comments make R sound like the greatest thing since sliced bread. I used it and found it that I did not like several features. Despite being a `lisp-like' language, it is really hard to use R to do any data manipulation. For example, how do you add a new line of data to a dataset? Can you scan a dataset, applying a function to each row - meaning getting inputs (username password ...) to your function? I wanted to create a new `derived' dataset for analysis by filtering the original data. It seemed impossible. Confusion about "orthogonality": to get histogram data, you are advised to call hist(dataframe, plot = false). Obscure names for functions! Undocumented interfaces. Is a dataset inherently rows or columns, and why do you do data[ 1, ] to get a row instead of data[ 1 ]? Array indexing seemed quite confusing to me. I am glad that the authors understood that continuations were important. I wish that they had been able to understand the rest of the Scheme lesson that you build your special language ON TOP of a fully featured language like Scheme. I much prefer Mathematica's decision to use GreatBigLongNamesThatYouCanRemember. Surely the same idea would be useful in the land of statistical tests. (But in this, the developers are constrained by their desire to implement S... so I am really complaining about S.) My conclusion is that R is good at producing plots (although again, I think Mathematica has a MUCH cleaner handling of this - especially plot options!), and I'm sure it does great statistics, but integrated solutions are not possible - do everything step by step, and by hand. 2004-09-23 23:35:48.157 2004-09-23 23:35:48.141