?

Log in

No account? Create an account
Decision Analysis seminar all day today. It's neat. Tomorrow we're… - The Mad Schemes of Dr. Tectonic [entries|archive|friends|userinfo]
Beemer

[ userinfo | livejournal userinfo ]
[ archive | journal archive ]

[Jan. 26th, 2006|11:00 pm]
Beemer
Decision Analysis seminar all day today. It's neat. Tomorrow we're doing a case study about a fictional post-hurricane levee-repairing plan.

Mel & Kate came over for Cheapass games tonight (Deadwood). It was fun; Mel won. I forgot that one of the better strategies is just to go for the low-paying but frequent options.

So tired. Bed now.
LinkReply

Comments:
[User Picture]From: melted_snowball
2006-01-26 09:49 pm (UTC)
Isn't this just nonlinear stochastic optimization?
(Reply) (Thread)
[User Picture]From: dr_tectonic
2006-01-27 05:54 am (UTC)
Which, the decision analysis, or the getting paid scale in Deadwood?
(Reply) (Parent) (Thread)
[User Picture]From: ng_nighthawk
2006-01-27 06:07 am (UTC)
*in deep, profound tone*

Maybe. . . maybe it's both.
(Reply) (Parent) (Thread)
[User Picture]From: melted_snowball
2006-01-27 08:13 am (UTC)
Thank you. After a day like yesterday, I needed that laugh.

I more meant decision analysis. It's been a while since I played a Cheapass game. (I remember the anagram one about getting Napoleon off Elba right now...)
(Reply) (Parent) (Thread)
[User Picture]From: dr_tectonic
2006-01-27 04:06 pm (UTC)
Um, maybe? It depends on what you mean by "nonlinear stochastic optimization". I suspect the answer is "partly".
(Reply) (Parent) (Thread)
[User Picture]From: melted_snowball
2006-01-27 07:19 pm (UTC)
Mostly something like "pick values of decision variables x_i so as to maximize E[f(x)], while satisfying potentially confusing constraints on the x"

[Basically, pick feasible values of x so as to maximize expecetd utility.]
(Reply) (Parent) (Thread)
[User Picture]From: dr_tectonic
2006-01-27 08:08 pm (UTC)
Ah! The answer is, indeed, "partly".

It's like, "find the x_i, construct f(x), determine which x_i matter, maximize E[f(x)], find potentially useful VoPI, VoII, and VoC (value of perfect info, imperfect info, and control) values, and relate the results in a suitable form depending on risk tolerance, risk preference, and decision-making style".
(Reply) (Parent) (Thread)