Talk:Variable structure

Dependencies instead of causality

How to read discussions

Fact discussion: .
Opening statement: The name of the causality attribute should be dependencies. Closing statement: Accepted. (A closing statement, when resolved, should be updated to the main page.)
Argumentation: ←--1: . Dependencies is more clear because it points to variables that affect this variable; causality may also mean downstream causalities. Promoted by Erkki Kuusisto, written down by --Jouni 10:26, 19 October 2008 (EEST) (type: truth; paradigms: science: defence)

Key variable

"Key variable" as a term sounds ambiguous and very broad compared to the definition given in the text. Perhaps this type of variable should more appropriately be called a "diagnostic variable"? -- Erkki Kuusisto 16:43, 22 January 2008 (EET)
←--1: . From my point of view a key variable is actually simply an indicator. So would be "diagnostic variable", which as a term sounds neither familiar nor clarifying to me. For simplicity, I suggest discontinue using term "key variable", but instead using indicator or indicator variable as its replacement. --Anna Karjalainen 14:35, 24 April 2008 (EEST) (type: truth; paradigms: science: defence)

Discussion on formula attribute

How to read discussions

Fact discussion: .
Opening statement: 'Formula' is not an attribute of a variable. It is rather a process that links different variables?Needs editing Closing statement: Not accepted. (A closing statement, when resolved, should be updated to the main page.)
Argumentation: ←--1: . This issue has been discussed before, but I am not sure I agree with the fact that each variable has the attribute 'formula' in the description. I see it as follows (see also the full chain approach diagram): There are variables that describe reality, for example 'air pollution levels' or 'number of cardiovascular disease cases'. These variables have as attributes all the ones listed on the article page, but (as I see it) NOT a formula. Because there is not "one formula" for a variable. For example, transport can emit various pollutants as well as to noise. All of those pollutants as well as noise can lead to cardiovascular disease. Each pathway (from the exposure to cardiovascular disease, arrows in the diagram) has a different formula. This can never all be included in the definition of 'cardiovascular disease'. As I see it, a variable is more or less fixed, how to get there is not. Instead, the processes linking the variables should be separately defined, with a different set of attributes. You could also call these processes 'variables' if you want (although I think this makes it complex), but then they are different types of variables with different attributes. The processes are where to put the models and the formulas leading to the variables. Conclusion: I think it makes it a lot clearer to separate variables and processes. What do you think? (Anne) (type: truth; paradigms: science: defence) ⇤--5: . Anne is right saying that there is not only one path to a variable, e.g. cardiovascular disease incidence in a population. ALL of them should in theory be described in the formula, but in practice only the most relevant are dealt with. Also, there may be several alternative formulas to compute the same incidence. Again, all valid formulas should be included. And thirdly, Anne is right in saying that a formula is actually a process rather than a product. However, it is a process that is practically always needed when the result of a variable is derived. Therefore, it has its place as a subattribute. --Jouni 23:21, 13 July 2009 (EEST) (type: truth; paradigms: science: attack) ----6: . The calculation process described in the Formula may be of a general kind and worth describing independently from a variable. Then it becomes a method which is a process description. If this is done, the only thing that is needed in the Formula of the variable is a link to the method and a link to the input data used in the method. --Jouni 23:21, 13 July 2009 (EEST) (type: truth; paradigms: science: comment) ----2: . Variable is probably generally primarily perceived as an object of non-event type. Reality also includes objects of event type. These are phenomena that mediate the changes in the context (exterior) to the non-event(s) and manifest themselves as changes in the state of the non-event(s). The event and non-event type objects could be seen as different sides of the same coin, the other type not existing without the other. Or in other words, the non-event is a manifestation of the effect that the event causes in the prevailing context. The event and non-event objects can also be handled as composite objects that are composed of both event and non-event type objects. What is meant with the word variable in this approach is a composite object. It has attributes that mainly define the identity of the variable specifying its relations with its context(exterior), the result attribute, is an expression of the state of the variable, given the context and its relations of the variable, and the definition attribute describes the event(s) that manipulate the variable in ways that are then manifested in the state of the variable, i.e. its result. Whether a description of reality (e.g. a risk assessment) should be based on describing objects as composites or events and non-events separately is an interesting question and should be further considered. --Mikko 13:44, 30 July 2007 (EEST) (type: truth; paradigms: science: comment) ⇤--3: . After some reading and thinking, here is an attempt to clarify and correct the previous comment and to take the consideration a bit further. There are two primitive universal classes of objects: event and medium (I referred to the latter as non-event in previous comment). Events are phenomena that take place in the media, provided that the structures and the states of the media are favourable for the phenomena to occur, and whose realization can be recognized as changes in the structures and the states of the media. Events can be (hierarchically, vertically) divided into sub-events and similarly media as sub-media. Events and media can also be perceived as event-medium composites, where the event and medium are interlinked (horizontally, across the same level of hierarchy) by causal relations. Several sub-events can occur in a single medium and a single event can occur in several sub-media. Events and media compose chains or networks of phenomena, where events interact with their media, which further interact with other events which again interact with other media. An event and its media are inherently bound and the effect of an event can only be observed through changes in (the structure and state of) its media, so these chains are probably most comfortably perceived as comprising of event-medium composites which are linked to each other causally ([event-medium]→[event-medium] etc.). What is referred to as variable in the approach described on the article page, is actually an object of event-medium composite kind. Its attribute name is an identifier of the object, attributes scope and definition describe its structure (hierarchical and horizontal relations), and to some extent its purpose, while the attribute result describes its state with the help of the attribute unit. The descriptions of a variable thus include both descriptions of the medium and the event(s) that are defined to be included within the boundary (scope) of the variable. I suggest that variable, as an object of even-medium composite type, is chosen as the abstraction level of observation, but that the structure descriptions of the variables be made more explicit in what events and media are included within the variable and what are the causal relations between them as well as the causal relations to other variables. Note: These considerations are strongly influenced by a universally applicable PSSP ontology, described in e.g. Pohjola V.J.: Ontological Approach to Formalize Waste Management, http://www.springerlink.com/content/r1n10r828nt57622/. --Mikko 16:36, 31 July 2007 (EEST) (type: truth; paradigms: science: attack) ----4: . Mikko, could you "translate" this to the example made by Anne? I admit, I don't really understand what you want to say. Or, maybe, I am not able to transfer the abstract into usability. Anne has asked what happens if you have a variable "number of cardiovascular disease cases" which is cause by several pollutants/stressors. Would "number of cardiovascular disease cases" be something like a class containing variables like "number of cardiovascular disease cases due to PM" and "number of cardiovascular disease cases due to noise"? How would that take into account the combined effect of PM and noise? (if we have an ERF for noise AND PM e.g.? --Alexandra Kuhn 09:22, 7 March 2008 (EET) (type: truth; paradigms: science: comment)

How to read discussions

Fact discussion: .
Opening statement: MC is not an uncertainty dimension of a variable Closing statement: Accepted. (A closing statement, when resolved, should be updated to the main page.)
Argumentation: ----1: . I don't know if I wrote correctly what Jouni was saying. --Alexandra Kuhn 08:06, 10 June 2008 (EEST) (type: truth; paradigms: science: comment) ----2: . This sounds like EVERY variable should perform a Monte Carlo analysis (at least those with numerical results). Does it make sense??? --Alexandra Kuhn 08:06, 10 June 2008 (EEST) (type: truth; paradigms: science: comment) ----3: . From WHAT do you take random samples? --Alexandra Kuhn 08:06, 10 June 2008 (EEST) (type: truth; paradigms: science: comment) ←--4: . You are correct. Some variables may be deterministic, i.e. constants. In addition, taking a sample is not the only way to describe a distribution. But if the variable is described as a distribution, then the sample means that N values is randomly drawn from the distribution. These N values then is the sample. --Jouni 00:27, 14 January 2009 (EET) (type: truth; paradigms: science: defence)

Decision variables -- Anne.knol 17:13, 20 February 2008 (EET)

How to read discussions

Fact discussion: .
Opening statement: Decision variables have different (or less) attributes than regular variables Closing statement: Not accepted. (A closing statement, when resolved, should be updated to the main page.)
Argumentation: ←--1: . Decision variables cannot be quantified, since they describe decisions or scenarios. They therefore don’t have a formula or unit, or any up- or downstream variables. Is a decision variable actually a variable at all…? --Anne.knol 17:24, 20 February 2008 (EET) (type: truth; paradigms: science: defence) ⇤--2: . The scope of a decision variable is: ”Which are plausible decision options in this situation (description of the situation)?) The definition contains information about how you know which options actually exist and are plausible in the defined situation. The result lists all plausible policy options. It should be noticed that the result of the decision variable does not necessarily match the decision scenarios of a particular assessment. --Jouni 17:55, 20 February 2008 (EET) (type: truth; paradigms: science: attack) ----3: . What about the unit and formula? --Anne.knol 17:57, 20 February 2008 (EET) (type: truth; paradigms: science: comment) ----4: . Unit and formula are irrelevant and left empty in this case, because the result is not numerical but a list of options. --Jouni 18:04, 20 February 2008 (EET) (type: truth; paradigms: science: comment) ----6: . However, even if some variables are qualitative and do not have Unit, this is not a strong argument to deviate from the principle that all variables have the same structure. Some variables just don't need all subattributes. But they all need all main attributes. --Jouni 13:25, 10 September 2008 (EEST) (type: truth; paradigms: science: comment) ----5: . By the way, this discussion should be moved to Talk:Variable. --Jouni 18:04, 20 February 2008 (EET) (type: truth; paradigms: science: comment)