Decision Tree: Issue with cost about radiant

[vnijs]

Can you share an example? 'cost' is currently only for use in decision nodes (e.g., buy a research study and pay $X).

[vnijs]

Here is a simple example with a cost. An agent costs $1000, is it worth it given the difference in payoffs?

name: Cost example
type: decision
Hire agent:
    cost: 1000
    type: chance
    Low pay:
        p: 0.3
        payoff: 2000
    High pay:
        p: 0.7
        payoff: 8000
No agent:
    type: chance
    Low pay:
        p: 0.7
        payoff: 2000
    High pay:
        p: 0.3
        payoff: 8000

cost_example.pdf

[ezriah]

Here's the decision tree I'm working on. The later nodes don't inherit the
cost:

Name: Survey
type: decision
Survey:
cost: 30
type: chance
Unfavorable:
p: .7
payoff: 0
type: decision
Drill:
cost: 100
type: chance
oil:
p: .143
payoff: 800
dry:
p: .857
payoff: 0
Sell:
payoff: 90
Favorable:
p: .3
payoff: 0
type: decision
Drill:
cost: 100
type: chance
oil:
p: .5
payoff: 800
dry:
p: .5
payoff: 0
Sell:
payoff: 90
NoSurvey:
type: decision
Drill:
cost: 100
type: chance
oil:
p: .25
payoff: 800
dry:
p: .75
payoff: 0
Sell:
payoff: 90

On Wed, Nov 18, 2015 at 4:22 PM, Vincent Nijs [email protected]
wrote:

Can you share an example? 'cost' is currently only for use in decision
nodes (e.g., buy a research study and pay $X).

—
Reply to this email directly or view it on GitHub
#41 (comment).

[ezriah]

In this example, the payoff for hiring an agent with a cost of 1000 and a
payoff of 2000 should ultimately result in a payoff of 1000 (2000-1000).

On Wed, Nov 18, 2015 at 4:30 PM, Vincent Nijs [email protected]
wrote:

Here is a simple example with a cost. An agent costs $1000, is it worth it
given the difference in payoffs?

name: Cost exampletype: decisionHire agent:
cost: 1000
type: chance
Low pay:
p: 0.3
payoff: 2000
High pay:
p: 0.7
payoff: 8000No agent:
type: chance
Low pay:
p: 0.7
payoff: 2000
High pay:
p: 0.3
payoff: 8000

—
Reply to this email directly or view it on GitHub
#41 (comment).

[vnijs]

That is not how it works (i.e., later nodes do indeed not 'inherit' cost). Rather, the cost is taken into account appropriately when you fold back the tree. I think it would be confusing if you put in a payoff for a terminal of say, 2000, but the tree shows 1000. That said, it would be nice to see where a cost is deducted. There are something in-the-works that will address that.

In sum, the tree is correct. It is just not displayed as had you expected. In fact, I have not seen tools that display output as you suggest. If you have seen such examples please let me know.

[ezriah]

My operations research textbook (Introduction to Operations Research by
Hillier and Lieberman) teaches it that way. They use the same example but
perform analysis with Analytics Solver Platform, an Excel plugin, which
gives the results in the terminal nodes as the payoff minus the cumulative
costs. Thanks for such a quick response. Let me know if you'd like more
details.

On Wed, Nov 18, 2015 at 5:32 PM, Vincent Nijs [email protected]
wrote:

That is not how it works (i.e., later nodes do indeed not 'inherit' cost).
Rather, the cost is taken into account appropriately when you fold back the
tree. I think it would be confusing if you put in a payoff for a terminal
of say, 2000, but the tree shows 1000. That said, it would be nice to see
where a cost is deducted. There are something in-the-works that will
address that.

In sum, the tree is correct. It is just not displayed as had you expected.
In fact, I have not seen tools that display output as you suggest. If you
have seen such examples please let me know.

—
Reply to this email directly or view it on GitHub
#41 (comment).

[vnijs]

Unfortunately I don't have the book and don't know the plugin. I expect the plugin does not, however, automatically update the payoff for the terminal nodes. I assume the user provides the final payoff, adjusted for cost. If that is not a correct assumption could you perhaps send me the xls file and I'll look up the plugin.