Log in

No account? Create an account

Mon, Jan. 29th, 2007, 11:37 am
jpfed: LifeStat

I realized earlier last week that LifeStat is just a particular combination of services that could communicate with Double-X's central infrastructure. Specifically, it is two different views on time-series data (a calendar view and a cross-correlation view) and an extension that would allow over-the-network merging of files. Designing the right architecture for Double-X's backbone would allow LifeStat to just be a particular configuration of Double-X.

That said, if Double-X were extremely configurable, then it would be good to have an app that wires together the backbone and services to produce custom solutions so that end-users (for whom the full power of Double-X may be irrelevant) are presented with exactly and only what they want to see in their situation. The limit as x goes to infinity is that we program the backbone and relevant services AND pre-package them together to provide solutions for the end-user without the end-user being at all aware that they are using a specialized form of a much more general package.

Sun, Jul. 23rd, 2006, 11:58 pm
jpfed: Allow the user to edit what they're looking at.

I had some ideas today.

1. Statisticians are interested in patterns that show up in the data, but they are wary of patterns that have showed up by coincidence. They have ways of estimating the probability that a pattern has showed up by chance vs. the probability that the relationship actually exists. Most of the time, we're interested in relationships that are 95% likely NOT to have been coincidental.

A correlation table where each correlation is shown as a filled rectangle of solid color (with the colors depending on the strength of the relationship) is a nice start. But it would be even better if relationships meeting the 95% criterion were also given drop shadows. That would make them really pop out, even if color wasn't enough.

2. The correlation table should have an editable vertical gradient strip running down its right side, so the colors it uses to represent relationships can be edited. My preferred color scheme would be terrible for colorblind people, and probably vice versa.

3. I like the idea of being able to edit what I'm looking at. For example, imagine clicking through your bookmarks in Firefox. While navigating through Bookmarks-Myfolder-, you decide that you need to create MySubFolder. But you can't right-click in Bookmarks-MyFolder to create MySubFolder. In reality, if I'm in Bookmarks-MyFolder, and I decide that I want a new folder in there, I have to back out and go to Bookmarks-Manage Bookmarks... I want to imagine that the gui depicts a continuous space. That I can look at something but not be able to edit it is a discontinuity in gui-space. It's a minor complaint in this instance, but it brought to mind a greater principle that should be adhered to- you should be able to edit what you're looking at.

So the moral of the story is, if you get somewhere by navigating there, then it would be nice to be able to edit/change what is displayed in that place once you're there. Sure, there might be a big, badass Preferences dialog with a million tabs in it accessible by the Edit menu. But if a tab in that preferences dialog is especially relevant to one area of the gui-space, then when you're in that area of gui-space the context menu should have the option of invoking the prefs dialog with that relevant tab pre-selected.

Thu, Dec. 31st, 2015, 11:59 pm
jpfed: Alright kids, let's begin

If you haven't read the uplifting mission statement in this community's info page, do so now. Your life will be changed forever.

Now that you've read that, we can go into some ground rules...Collapse )

  • general: not work-related
  • project: work-related, only stuff that's relevant to everybody
  • pluginapi: the plugin API (most work on specific plugins does not necessarily raise plugin API questions)
  • fileio: file formats/ import and export (probable child tags: excel, spss)
  • ui: user interface (probable child tags: frame, views, lnf)
  • history: data versioning/ undo on steroids (probable child tags: htree, scripting, summaries, annotation)
  • functions: analysis functions like ANOVA, t-test, correlations, etc. (probable child tags: scripting, math)
  • helpfile: work on developing the help system