The old laptop got some new life last month, with a makeover, and open source is a good part of it. I’ve been looking at open source solutions and championing them at work when possible (which is pretty much a losing cause unless you’re at the top of the org chart), and use the open source WordPress software for the blog, so it’s not a new idea for me.

I had intended to go whole-hog open source and run Ubuntu for the operating system on the laptop, but I ran into some problems that I didn’t want to spend quite that much time solving. Hardware incompatibility was the show stopper for the kid. Heck, much of the hardware on this laptop isn’t compatible with base Windows XP, having needed driver downloads from the manufacturer and some tweaking to work right to begin with. The Ubuntu short story for me was that I could have gotten it to work perfectly with my hardware if I could have gotten an internet connection, but the hardware that wouldn’t work with Ubuntu out of the box included the modem, the broadband card, and the networking cards. I had another Windows machine that had a good internet connection, but the “sudo” command-line internet wizardry would have to come from the Ubuntu machine. At the end of several frustrating hours over a few evenings, I decided that Ubuntu and I would have to wait for our merger until I needed a new machine – which would be ordered with compatible hardware from the get-go. The current hardware is old but not ancient, and sufficient with a stripped install of XP.

Some (but not all) of the stuff I’m using includes:

Filezilla
Firefox
Thunderbird
7-Zip
BoncEnc
InfraRecorder
Audacity
MP3 Gain
FLV Player

Many of these are being used to rip old cassettes to MP3 format and level the sound, because dammit, I already paid for Bat Out Of Hell (along with hundreds of other albums), I just want to hear it (them) on my new SUV’s speakers, and my gas-guzzler doesn’t play cassettes. Not to mention that listening to the entirety of George Thorogood’s concert in Cincinnati is like using a dirty gas station toilet without having a newspaper to read - but there are three songs I like on that tape, that deserve rotation.

My favorite open source item, however, is OpenOffice.

I’m most happy with OO Calc and some of the additional functionality from plugins like OOoStat (which uses macros in OO Basic to mimic the functionalities of several Excel Add-Ins), and I’ll be digging into accessing R through OO Calc plugins before the month is out. I’ve been testing out the analysis and charting functionality, and while it does some things differently, it does all the same things as Excel (so far), with some workarounds. My biggest gripe is not having the keystroke shortcuts out of the box. I know that I can customize the keystrokes and re-create them, but I’ve decided to learn the new patterns in OO and use it exclusively from home, if I can.

Here’s some OOoStat simple multivariate regression output from a combination of U.S. Census data for 1837 zips with valid entries in all categories of interest, vs. zip-code level data from the TX Department of Insurance. I’m modeling BIPD pure premiums here; note that it’s a personal test analysis and not anything I’m doing for “work,” and it’s publicly available data only. I chose it for its applicability to the test process. Click any table or chart for a better view.

Here’s a graph of the fitted vs. actual pure premium for a similar analysis, limited to the 500 zips that had the highest amount of earned exposures. Note that it’s a personal test analysis and not anything I’m doing for “work,” and it’s publicly available data only. I chose it for its applicability to the test process.

Here’s the most powerful predictor of the Census dataset I applied, population per square mile of land area, treated as follows: 20 equal-weighted groupings by population, with a log function applied to the density number. Graph is the simple average of the lower and upper ranges of the log function on the x-axis, with weighted-average population density and BIPD pure premiums on the y-axes. Note that it’s a personal test analysis and not anything I’m doing for “work,” and it’s publicly available data only. I chose it for its applicability to the test process.

I haven’t tried out OO Base in full yet, although I’ve done a few simple reports in it. The real tests will be when I hit OO Base with some public data that I’ve currently got canned in Access MDBs. I’ll probably be doing clustering analyses with R when I get it running.

Bottom line, I’m happy with the capabilities of the open source software I’ve been playing with at the house, and I plan to continue moving in that direction as time goes by.