Syntax Highlighter Test Post

I am trying to figure out how to use three things:

1. Alex Gorbatchev's SyntaxHighlighter
2. Yihui Xie's custom R-Brush for highlighting R
3. Google Codes' website for hosting the javaScripts

test.func <- function(r) {
    return(pi * r^2)
}
test.func(1)

It is proving difficult so far.

So I got it using Alex's code direct from his website. I have tried changing the starting line number and highlighting a single line. Also I added a grey border on the expanded box and got a title to appear.

I think I should be able to get the code to pop out into its own plain text box, making it easy to copy and paste into the R editor. I gather this has been deprecated and you use a simple double-click, copy - paste.

Next challenge is to get it from the google-code website so I can design my own color scheme!

Well, so far I have the javascripts running from google-code, but every time I add the cascading style sheets, it stops working - I bet it is something to do with relative directory structures.

No, not directory, I think it MIME type. Apparently Google code serves up the file as plain-text not as CSS, so my template doesn't interpret it correctly. Trying to figure out how to set the MIME type correctly on Google code.

I tried using Google Docs - didn't work. So, after down loading Tortoise Subversion client and using it to upload files with the correct mime settings, everything is being driven off Google-Code. However, I get the strong impression that Google is not happy about this use of Google Code - I have to look into their suggestions for hosting javascript, css, etc.

Ch-Ch-Ch-Ch-Changes

I guess I am showing my age with that title!

In my previous post I looked at the problem of detecting a change in a randomly generated time series. At some point in the time series, the mean and/or the standard deviation used to generate the series makes a step-change.

In this post I want to look at a more realistic problem of multiple changes at unknown times in the time series. I will make available the R-script for anyone who wants to play with this themselves - it is the best way to get a feel for the algorithm I plan on presenting. Finally I will apply the algorithm to the monthly returns data for the last 14 years for the Winton Diversified Fund.

Changes

I have been working on the math behind my manager-monitoring services with the objective of making them more rigorous. Rather than just looking at the statistics and spotting anomalies, I want to be able to pick them out objectively. I thought I would share some of the basics of monitoring for change in a couple or three posts. First I will review off-line change detection for a single change, then look at multiple changes and finally extend that to on-line monitoring.

Off-line monitoring means we have the entire data series and we are looking for the point or points in the history where some aspect of the data (such as the mean or variance) changed in a statistically significant way. On-line monitoring is similar but is carried out in real-time as the data arrives. For example: Based on today's return data, can we now conclude, based on the preponderance of the evidence, that a manager's performance has changed compared to some baseline?

In this post I will start out with a statement of the problem illustrated by the historical performance of Winton's Diversified Program. I will give some background with a quick refresher on Maximum Likelihood Estimation (MLE) and its extension to the Generalized Likelihood Ratio (GLR). I am going to explore the problem of detecting a single change at an unknown time in the mean and/or standard deviation of a Gaussian series. I will show how using R's probability density functions make the math a lot easier.