search

stat157 / background

0 stars 4 forks source link

Data Domains #30

Open rerock opened 10 years ago

rerock commented 10 years ago

After reading more on Luen, Stark's paper, I think they set their data domains as: Use an earthquake of magnitude 5.5 or greater occurs anywhere in the world from 2000-2004, to predict that an earthquake at least as large will occur within 21 days and within an epicentral distance of 50 km.

So probably magnitude is something we need to take into consideration. I tried with all 1938-2013 SCEC earthquake's Mag >4.5, and now we have a data set with 640 points, which is roughly 1/6 of iran data. (Iran data takes 2 mins to be finished running). I have uploaded the New Data With Mag >4.5 csv file here. The new data frames with Mag >4.5 for ETAS & SAPP packages are uploaded as well. These Data frames for ETAS & SAPP has only 250 data points so you can play around with them first.

Sorry if I am confusing you, I got to go to a meeting in 15 mins. I can explain more after my meeting at 11pm. Comments and concerns are welcome.

alexchaomander commented 10 years ago

Do you know how many earthquakes is "all earthquakes in the world from 2000-2004" for Luen's paper? That sounds like it could still be quite a large number of data points.

Although if we just limit our scope to 640 points, then it should be more manageable.

Thanks for going through the paper!

rerock commented 10 years ago

As far as I understand, they didnt say how many earthquakes in the world from 2000-2004. They didnt say the number of earthquakes happened with mag>5.5, either.

If we want to have a larger set of data, mag>4 of SCEC will give us 2027 data points.

Another question is should we look at more than just SCEC data? I guess we dont have an answer until we start playing around with the model. Thanks for the reply & please pass the information to your group so that they can have a more defined data set.