Code organziation during exploratory data analysis?

Hello,

I'm interested in how you organize your code for exploratory/preliminary work with complex datasets. Imagine a case where you have some data in the workspace and want to manipulate and plot parts of it, but there's no clear workflow yet -- you're making up your analysis strategy as you go.

For this kind of work, I typically use scripts. I load my raw data into the workspace and then manipulate and plot it using scripts with many sections.

I like how flexible and self-contained scripts are, but it gets complicated when I want to re-use short blocks of code, e.g. for analyzing several variants of my data in parallel. To re-use code, I see two options: anonymous functions and m-file functions. What I'm missing is something in between, namely nested functions within scripts. Often, the code blocks that I want to re-use are too complex to be converted to an anonymous function. But I might have dozens of them in the same script, and I don't want to maintain a mess of function files that are quickly getting out-dated and are more cumbersome to edit than code that's directly within the script file.

So how do you handle this situation?

• Do you copy/paste code and edit it manually?
• Do you create many little function files?
• Am I wrong using scripts in the first place?
• How does MathWorks envision work like this to be done?

Thanks, Matthias