Result range: Difference between revisions
m (Domain moved to Result domain: See discussion 'Renaming needed?' on Talk page.) |
(small edites here and there) |
||
| Line 2: | Line 2: | ||
[[Category:Glossary term]] | [[Category:Glossary term]] | ||
<section begin=glossary /> | <section begin=glossary /> | ||
:The '''domain of a variable''' contains all possible values that the result of the variable may get with probability P>0. The domain is restricted by logical (the number of actual individuals must be an integer), rational (concentration cannot be negative or above 1000 g/kg), and observational ([[:en:TCDD|TCDD]] concentration in humans has never been observed to be above 144000 ng/kg fat) statements. | :The '''result domain of a variable''' contains all possible values that the result of the variable may get with probability P>0. The result domain is restricted by logical (the number of actual individuals must be an integer), rational (concentration cannot be negative or above 1000 g/kg), and observational ([[:en:TCDD|TCDD]] concentration in humans has never been observed to be above 144000 ng/kg fat) statements. The utility of result domain comes from the dependencies to other variables. Child variables are probability distributions conditional on their parents. If the result domain of the parents is defined and more or less unchanged, it facilitates the definition process of the children. The conditional probability distribution of a child is only defined for the parent values within its result domain. | ||
<section end=glossary /> | <section end=glossary /> | ||
Should ' | Should 'Result domain' be a sub-attribute of 'Definition' in a variable? Alternative solution: it is discussed under 'Data', because in practice, the result domain is based on Data. Third alternative: it is put under 'Result', maybe as a sub-attribute; this is not a good idea, because the result domain has implications on e.g. 'Formula', and it is not good if the result affects Definition. | ||
:→ Because clear understanding does not yet exist, the attribute structure is not changed, and the domain is described under 'Data'. {{ | :→ Because clear understanding does not yet exist, the attribute structure is not changed, and the result domain is described under 'Data'. {{reslink|Renaming needed?}} | ||
==Scope== | ==Scope== | ||
How can a plausible domain for a variable be found? What are the criteria for inclusion or exclusion of a particular value or range? | How can a plausible result domain for a variable be found? What are the criteria for inclusion or exclusion of a particular value or range? | ||
==Definition== | ==Definition== | ||
| Line 16: | Line 16: | ||
===Rationale=== | ===Rationale=== | ||
In expert judgement, it is a common practice to define the domain by calculating the difference between the fractiles 0.05 and 0.95, extend that by k % (k is often 10 % or found by optimising), and subtract it from and add it to the fractiles 0.05 and 0.95, respectively. | In expert judgement, it is a common practice to define the result domain by calculating the difference between the fractiles 0.05 and 0.95, extend that by k % (k is often 10 % or found by optimising), and subtract it from and add it to the fractiles 0.05 and 0.95, respectively. | ||
In [[open assessment]], all values are included in the domain by default. A suggested procedure is to exclude any range X that passes a critical [[argumentation]] about the exclusion. The argumentation has a statement of the following form: "It is impossible that the result of the variable may be within range X." If the statement is accepted, X is excluded from the domain. {{ | In [[open assessment]], all values are included in the result domain by default. A suggested procedure is to exclude any range X that passes a critical [[argumentation]] about the exclusion. The argumentation has a statement of the following form: "It is impossible that the result of the variable may be within range X." If the statement is accepted, X is excluded from the result domain. {{reslink|Should inclusion or exclusion be the default?}} | ||
==Procedure== | ==Procedure== | ||
All values are included in the domain by default. Any range X that passes a critical [[argumentation]] about the exclusion will be excluded. The argumentation has a statement of the following form: "It is impossible that the result of the variable may be within range X." If the statement is accepted, X is excluded from the domain. | All values are included in the result domain by default. Any range X that passes a critical [[argumentation]] about the exclusion will be excluded. The argumentation has a statement of the following form: "It is impossible that the result of the variable may be within range X." If the statement is accepted, X is excluded from the result domain. | ||
==See also== | ==See also== | ||
* [[Variable]] | * [[Variable]] | ||
Revision as of 06:50, 29 September 2008
| [show] |
|---|
<section begin=glossary />
- The result domain of a variable contains all possible values that the result of the variable may get with probability P>0. The result domain is restricted by logical (the number of actual individuals must be an integer), rational (concentration cannot be negative or above 1000 g/kg), and observational (TCDD concentration in humans has never been observed to be above 144000 ng/kg fat) statements. The utility of result domain comes from the dependencies to other variables. Child variables are probability distributions conditional on their parents. If the result domain of the parents is defined and more or less unchanged, it facilitates the definition process of the children. The conditional probability distribution of a child is only defined for the parent values within its result domain.
<section end=glossary />
Should 'Result domain' be a sub-attribute of 'Definition' in a variable? Alternative solution: it is discussed under 'Data', because in practice, the result domain is based on Data. Third alternative: it is put under 'Result', maybe as a sub-attribute; this is not a good idea, because the result domain has implications on e.g. 'Formula', and it is not good if the result affects Definition.
- → Because clear understanding does not yet exist, the attribute structure is not changed, and the result domain is described under 'Data'. R↻
Scope
How can a plausible result domain for a variable be found? What are the criteria for inclusion or exclusion of a particular value or range?
Definition
Rationale
In expert judgement, it is a common practice to define the result domain by calculating the difference between the fractiles 0.05 and 0.95, extend that by k % (k is often 10 % or found by optimising), and subtract it from and add it to the fractiles 0.05 and 0.95, respectively.
In open assessment, all values are included in the result domain by default. A suggested procedure is to exclude any range X that passes a critical argumentation about the exclusion. The argumentation has a statement of the following form: "It is impossible that the result of the variable may be within range X." If the statement is accepted, X is excluded from the result domain. R↻
Procedure
All values are included in the result domain by default. Any range X that passes a critical argumentation about the exclusion will be excluded. The argumentation has a statement of the following form: "It is impossible that the result of the variable may be within range X." If the statement is accepted, X is excluded from the result domain.