Carl Boettiger

Morn­ing has been rather dif­fi­cult try­ing to get sim­u­la­tions up and run­ning.  Par­al­lel library isn’t run­ning on zero or one, vari­ety of errors that seemed to be machine depen­dent and rem­i­nis­cent of the ~./sfCluster prob­lem, 3 hrs of debug­ging revealed this was indeed just a bug in my updated code.  (In chang­ing to my object-oriented simulate.gauss and update.gauss func­tions rather than call­ing the spe­cific func­tions directly I hadn’t updated all the func­tion calls).

Updated Runs

The com­par­i­son between tau curves com­pares sim­u­la­tions under the case with no loss in sta­bil­ity with the case of grad­ual loss (same ini­tial sta­bil­ity but adds a grad­ual loss rate), while the Monte Carlo test com­pares sim­u­la­tions between mod­els that have both been esti­mated from the same data set.  The lat­ter two mod­els are more sim­i­lar, as esti­mates of con­stant rate put that con­stant near the aver­age, whereas the Tau sim­u­la­tions leave it at the ini­tial value of what it assumes in the chang­ing rate model.  (That is, one has the model for the sta­bil­ity para­me­ter \(R(t) = R_0\) while the other has \(R(t) = R_0 + m t\), and in the Tau com­par­i­son the \(R_o\) val­ues agree, whereas in the esti­ma­tion they differ.  )

WriT­ing

Have con­tin­ued quite a bit of writ­ing on the man­u­script fol­lowed by a very good meet­ing with Alan.

Trou­ble with Ama­zon AMIs

Some trou­ble with Ama­zon com­put­ing I still haven’t solved.  I go to launch my ama­zon cloud using my exist­ing AMIs and they don’t exist

I go to re-add my ama­zon AMI using

ec2-register cboettig-private-bucket/image.manifest.xml -n cboettig-ami -C ~/.ec2/cert-H....pem -K ~/.ec2/pk-H...pem --region us-west-1

and I get error

Invalid or unaccessible kernel id in ami manifest:

Will have to post on the devel­oper forum to see if I can learn what is going on, does not seem that AWS sup­port cov­ers tech­ni­cal questions.