12. MAKING A LIST OF RELATIONSHIPS (KINESES)

The next step is to connect the variables together. But before that can be done, there must be some agreement on what is meant by “connecting” or “relating” the variables. In COMPASS, “relating” is taken in a causal sense. In other words, Variable X causes a change in Variable Y. Different words that might be used instead of “causes” are “influences”, “affects”, “changes”, and just about any other word that signifies an action of X with regard to Y (for example, X “manages” Y). In COMPASS we use the phrase X “influences” Y or Y “is influenced by” X as a generality for all these situations. See the template in Appendix A.

What is not meant by X “influences” Y is that X is a definition for Y (or vice versa) or X is associated with Y in some analogical fashion. To illustrate the latter, one variable may be “National Security Problem” and another may be “Pet Security Problem”. They may have nothing whatsoever to do with each other except that they are analogous because both are security problems.

A preliminary task before specifying relationships is to copy and paste the following text:

            ------------------------------------------------------------------------------------------------------------

            Influencing:

            ------------------------------------------------------------------------------------------------------------

            Influenced:

            ------------------------------------------------------------------------------------------------------------

underneath any “further description” of each variable. The result should look like TP Page12.

Perhaps the easiest way to capture relationships is to start with the first variable on the list and ask if any of the other variables influence that one “directly” (meaning with no other variable being an intermediary). If the answer is “yes”, then do two things:

            (1) Add the ID Code and the brief description of the influencing variable (X) to the “Influencing” row of the influenced (Y) variable.

            (2) Add the ID Code and the brief description of an influenced variable (Y) to the “Influenced” row of the influencing (X) variable.

As an example, suppose in the listing of the two variables in TP Page12, you decide that the use of license plate recognition technology (TV-ANPR) leads to (influences) plate copying (LPCLONE). So, you would put the ID code TV-ANPR and corresponding brief description on the “Influencing” row underneath the variable LPCLONE and LPCLONE on the “Influenced” row under the variable TV-ANPR.  The result would look like TP Page13.

You now proceed to the second variable on the list and run through each other variable to see if it is an influence. And so on until you reach the bottom of the list. This is a kind of brute force approach which, with some experience, can be tempered with good judgment.

Several checks now have to be made to ensure things make sense:

            (1) Is there one variable (or more) that represents the “bottom line” or ultimate goal of the set of relationships? If not, identify such a variable.

            (2) Is each variable influencing, or influenced by, at least one other variable?  If not, then either connect the variable to another or drop it all together as not being relevant to the situation.

            (3)  Is there an influencing path between at least one variable and a given “bottom line” or “goal” variable? If not, put in the extra relationship(s) needed to create such a path.

            (4) Is there an influencing path from each non-goal variable to at least one goal variable?

An example of a set of variables and relationships that does not meet the criteria in checks (2) and (3) is portrayed in TP Page14, where the red box represents the (single) goal variable. In addition, the unattached variable near the lower left corner of the diagram illustrates the violation of check (1). The unconnected clusters throughout the diagram illustrate the violation of checks (3) and (4). Particularly noticeable is the fact that only two variables affect the goal variable (both directly, none indirectly through another variable).

Note: we have yet to find a case, except for reproductions of specific regulations, in which all of the variables are connected to at least one other as well as connected to the goal variable (s) to create the needed paths. Most literature (e.g., a journal article) is highly disconnected, leaving the reader to fill in the missing pieces. The point thus is that we are not expecting perfection via the four criteria above. In fact, in some cases it may only be possible to provide a list of variables.

A special point should be made about extracting relationships from written material. Usually we go a paragraph at a time, picking out variables and relating them. Then on to the next paragraph with the same but separate process. You have to keep a constant outlook, however, for connections to previous paragraphs, because they are rarely spelled out. And without such connections, you are left with a series of isolated complexes, a situation which violates check (3) and/or (4) above. Regardless, you will have a subsequent conundrum on hand because you will have to decide whether to put in your own connections (that is, try to outguess the author), or leave the complex as isolated.

By the way, we have found it useful to sketch little “influence diagrams” like that TP Page10 and in parts of TP Page14 so we can keep track of the relevant variables in each cluster and how they relate to each other.  Then it is easier to connect the clusters together later.

As another special point, it sometimes is tempting to put in a large number of relationships, based on the assumption (which probably has some truth to it) that everything eventually is connected to everything else. But some common sense is helpful here, else you will have a lot of work to do. On the other hand, it seems reasonable to have at least two influencing variables for each influenced variable, on the basis that the level of any variable is a function of balancing forces (of which there has to be at least two for a balance to occur). Note that we confess that many times we have not followed our own advice on this point.

The next step is to put in any desired information about the relationships. This could include the sign, timing, strength, and many other features. It may also be desirable to put in examples, references, parameter values, and goodness-of-fit statistics (where such apply and are available). In some instances it might even be appropriate to enter conflicting opinions about relationships, including ones which may proclaim that there is no relationship. Rarely are any of these done, however.

All this kind of information should be put directly underneath the section showing the influencing variables. See TP Page15 for an example.

FrontPage     Previous     Next