Older blog entries for statbanana (starting at number 11)

I think I am going to use Hart's approximation for several distributions, some of the formulas can achieve precision out to 20 decimal places, this is the best I have seen so far. I really want to try and stay away from any expansion derived approximations which associate the number of observations while still getting the precision I need.

Anna -n- Schey

Back to distributions. I took a crack at normal today and came up with the following options, I could use an asymptotic expansion, or a power series I found an interesting approximation by C Hastings, at least it was interesting until I tried it and it produced crap results. All and all I am in love with the handbook of mathematical functions, and things are going well. If anyone has had any luck with a fast precise algorithm pls. do share :)

In personal news (in case anyone cares) my bf just got back from a trip and bought me a stuffed penguin from FAO Schwarz it's kwel.

I am also going to NC for the Tech Circus it's nothing huge and technical but should be fun. Plus I will get to see some friends, mgalgoci, mark, timball etc.

* anna naps with penguin *
:-)

12 Aug 2002 (updated 12 Aug 2002 at 23:29 UTC) »

I am working on an implementation of the cell means model for single factor ANOVA. I think this will be more effective in that this model can be used again in regression and calculated much more quickly then a typical least squares computation.

Some analysis I did on test score data was presented today and for the most part it went well. The one complaint I had is that at the end of the presentation the facilitator stood up and reminded everyone that these are "just numbers" He went on to express how simple it is to pervert data into any result you desire. While the perversion aspect is certainly true, I have a hard time believing that the information will be taken seriously when presented as just numbers.

This is a commen theme, people rely heavily on the analysis I give them, with little or no thought to the theory that is behind it. I suppose some people go around ignoring multicollinearity and serial correlation but I am not one of them.


*whew*

UGH...SAS on VMS is not pretty, sassy yes, pretty no. After playing with the settings for a while I managed to make it slightly more bearable. I sent an email about getting RIT to extend it's site license to include the UNIX version of SAS, this would make me a much happier data enthusiast :-) If they don't I will kick them in the shin until they do :-P

going on a wine tour tomorrow...yummy

30 Jul 2002 (updated 30 Jul 2002 at 17:47 UTC) »

Keeping in mind the speed advantage of storing variable data in rows I can see why the library concept is so appealing for SAS. Not only can you quickly refer to the dataset internally, but you can also control the storage and order of observations in order to minimize side and processing time.

Why is it so hot im my office????

hrm. Why does RIT consider it ok to disable analog without telling anyone ????? why why why ????

Yeah! I got ANOVA done today, just the basics SSTR SSE SST, MSE, F ......ect. Progress is being made :-) the only issue is mimicking the results produced by SAS, since I am strictly following distribution and summation functions I'm hoping they match. Grilling tonight :-)

9 Jul 2002 (updated 12 Jul 2002 at 05:36 UTC) »

hmmm. I wonder why it is that sometimes smart ppl turn out to be jerks. You would think they would be smart enough to know better. and they are :-) In other news I got some killer sas books today woo hoo

ARRG! Why does R use a binomial approximation for an F distribution this is not only stupid but also incorrect. As F is my favorite distribution this irks me to no end.

Back to work today :-( I talked with a prof about my project and got the math dept.to donate a computer heh heh heh. Albert is not feeling well :-( Get better soon albert.

I verified several functions written for R today. Most of the base distributions look good chisquare, normal, F, geometric, t, binomial and all of the other favs :-P. I also finished putting up a new web page :-) it's nothing huge but at least it is up to date.

2 older entries...

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!