Tuesday, June 30, 2009

Not so clear

So, the results of the simulation are not so clear. There might be a problem on the GALFIT modeling or on the way parameters were defined ...

Labels:

posted by Cris Da Rocha @ 9:54 PM 0 comments

Monday, June 29, 2009

Good and bad news

Good news! Simulation is ready, done, over! eeeeeeee

Bad news! Its analysis and interpretation is not as straight as I would like it to be ...

Labels:

posted by Cris Da Rocha @ 10:23 PM 0 comments

Friday, June 26, 2009

Is it almost done?

So! GALAPAGOS finished its task. Now is multicomp who is playing, just making multi-component fit with GALFIT. That is running and should run during the weekend.

Finishing the last objects from the last cluster! Just the worst cases ahead.

Dentist was painful!

Labels: ,

posted by Cris Da Rocha @ 11:24 PM 0 comments

Thursday, June 25, 2009

It seems to go ...

Yep! Seems that the GALAPAGOS part of the simulation is running smoothly now and should be done by tomorrow.

Alto a third of the sample is already fitted in multi-components, a second third is running and the third part is the one running in GALAPAGOS.

If I have enough processors, it will be done by monday.

Also a nice MTB ride to test the new/old/renewed bike!! Needs some derailleurs and breaks tuning (cables stretch) and pedals are still not there. But the new cassete solved the chain jumping issue! I know, sports on Bikes'n'Coffee :-)) It's back alive!!

Tomorrow morning ... dentist, again. After that still 2 appointments to go ... ahhhhh

Labels: ,

posted by Cris Da Rocha @ 11:59 PM 0 comments

Wednesday, June 24, 2009

Some more jumps

Some more jumps on the simulation, but seems that now it goes. I should have results until the weekend!

Labels:

posted by Cris Da Rocha @ 11:44 PM 1 comments

Tuesday, June 23, 2009

Still jump starting ...

So, still jump starting my simulation, but seems my walk arounds have worked, so now it's running ... hopefully until the end!

Also fishing failures, now one more is about to go and one is left!

Labels:

posted by Cris Da Rocha @ 11:59 PM 0 comments

Monday, June 22, 2009

Crash and Burn ... bug fixing and jump starting ...

During the weekend I tried to make the simulation run as fast as possible, when I could use a larger number of processors. Well, turns out GALAPAGOS was not meant to run several images with a single object, but several (about a hundred not 20,000) with several objects ... so some issues appeared. Some were solved, some I'm still stuck with.

No new option for Delta_Chi^2. Unfortunately since the variation of Chi^2 is typically larger than the number of parameters ... it gives the same result as pure Chi^2 ...

Also saving failed models around.

Labels:

posted by Cris Da Rocha @ 11:59 PM 0 comments

Friday, June 19, 2009

Simulations

Designing the simulation to estimate the efficiency of the process of multimodeling.

Still happening.

Labels:

posted by Cris Da Rocha @ 10:58 PM 0 comments

Thursday, June 18, 2009

Today not much ...

Well, large fraction of the day involved in the dentist and also taking care of Matheus.

One more cluster gone, now there's 2 fields missing ...

The problem is the automatic definition of goodness of the fit of the models. It seems that I'll need to do some simulations to settle this question (and some others).

My MTB is almost ready!

Labels: ,

posted by Cris Da Rocha @ 11:59 PM 0 comments

Wednesday, June 17, 2009

Bored ... part 137459216

Bored and tired of checking models :-) There's still SOME to go. It will go.

Also the results from Delta_chi_squared need eye ball checking to know if it's what I need ... more eye-ball checking ... aaaaaaaaah :-)

Tomorrow dentist ... aaaaaahh :-)

Labels: ,

posted by Cris Da Rocha @ 11:59 PM 0 comments

Tuesday, June 16, 2009

Bored ...

There couldn't be anything more boring than checking the catastrophic failure of the models and fix them. The percentage is small but the number is quite large for doing it "by hand" ... anyway, there's still some to go.

By coincidence David Hogg posted on his science blog about the IMPRS Summer School on Statistical Inferences from Astrophysical Data what gave me "the light" of asking him about how to choose the best model.

The solution he gave me was a "Delta-chi-squared" (Delta-chi-squared = [chi-squared_A + npar_A] - [chi-squared_B + npar_B]) where a large positive number favors model "A", a large negative number favors model "B" and small numbers (either positive or negative) leave things suspicious. The spirit is the same of the F-test. Now I'll implement it and check the results.

Just reminding that actually David is the father of the idea of this blog, that he suggested me 4 year ago ... and the blog "pays back" :-))

Relaxing from the boring hand work finishing renewing my MTB ... it's getting cool!!!

Labels: , ,

posted by Cris Da Rocha @ 11:16 PM 0 comments

Thursday, June 11, 2009

One more done ...

So, one field more ... finish modeling the worst cases, but some are really lost.

Tomorrow another field ...

Labels:

posted by Cris Da Rocha @ 11:59 PM 0 comments

Wednesday, June 10, 2009

Still remodeling

Remodeling express of the more obviously failed models is done (I love fast computers), now inspecting the not-so-obviously failed ones. But before gotta produce some figures on what is done.

Labels:

posted by Cris Da Rocha @ 11:59 PM 0 comments

Tuesday, June 09, 2009

Remodeling express

Basically automatizing the most typical modeling failure and making it run express on the number cruncher.

My linux PC had a "thing" today. Suddenly I've got a "kernel panic" message and nothing else worked and when rebooting nothing would happen. Then even the rescue disk would mount properly but a message about an error with the libc that was not there. Inspecting, I realized the link which is libc.so.6 was pointing to a weird "lib" (yep, to a file called lib, that doesn't exist). Repairing the link made everything come back to normal, until the next "heart attack".

Also updated the MacOS and not fully happy with Safari 4, somethings got worse, it looks better, and basically Apples condition #1 ... tsc tsc tsc

Labels: ,

posted by Cris Da Rocha @ 11:59 PM 0 comments

Monday, June 08, 2009

Life goes on :-)

So, seems that no one had additional ideas on how to solve my "goodness of the fit" automatic indicator. So F-Test is implemented. Anyway, if I (or anyone) come with a new idea it's easy to implement.

Run for 2 of the clusters and result looks good. Now running to the other ones. Of course there are other issues, like failed modeling, that I'm dealing with it now.

Also revising Elif's paper. Some comments but there's a lot of work there.

Basically that ... and trying to return to my good shape ... physically and mentally!

Labels: ,

posted by Cris Da Rocha @ 11:59 PM 0 comments

Wednesday, June 03, 2009

Light at the end of the tunnel?

F-test may be the way to solve the "goodness of the fit indicator issue", any other ideas?

Labels:

posted by Cris Da Rocha @ 11:07 PM 0 comments

Tuesday, June 02, 2009

Statistical problems

"Simple" problem.

I have three possibilities to model a single object. Those models have different number of free-parameters. Reduced Chi^2 and residua are affected by this different amount of free-parameters in a way that systematically larger number of free-parameters, "better" fits, according to these indicators. Unfortunately this is not true, specially if we consider physical meanings.

Can anyone give me some light on that?? :-))))

Labels: ,

posted by Cris Da Rocha @ 11:42 PM 2 comments