Importance of factors

Previous  Top  Next

Each example can be described by different attributes. Synonymous the term "dimension" is applied. Dimensions can be relevant for reaching a target criterion and therefore be factors. Factors can loose their importance and turn into irrelevant attributes. Shifts are possible.
The relationship between a factor and the target criterion is called the relevance of the factor. Synonymous, the term importance is used. A dimension is important, if the absence of a specific value at a specific factor leads to unacceptable results of a variant, to failure. Important dimensions are called factors. Critically important dimensions can not be modified without effects on the success of the variants. Specific values must be present. At the same time this means that factors are unimportant if any value can be observed at variants with acceptable success. The absence of specific values is tolerable.
Prior knowledge of the analyst helps to select the factors to be analysed. This prior knowledge not only enables analysis, it restricts the results as factors may be omitted from analysis mistakenly. Formerly unimportant factors may have gained relevance without notice. Dimensions that are known to be important do not hit us by surprise. Those factors that have gained importance should be observed carefully. And just those factors are difficult to detect and to include in an analysis. This leads to the recommendation to include currently unimportant dimensions. Those unimportant attributes may show drastic gains in importance. The careful selection of factors to be analysed determines the success of an analysis to a large extend.

For marketing success of a company, the dimension "memorability of phone number" may be important but not the dimension "phone number".

If importance could not be measured in exakt numbers, two options would be left open

·

A sharp destinction between relevance and irrelevance

·

Using fuzzy descriptions like "little relevance"

The first option requires a causal model of the object to be analysed (or a similar form of mapping of the reality). For each state of the components of the model the relationships between the elements have to be analysed. Following all causal paths, relevant and irrelevant elements can be distinguished.
The type of representation which sharply differentiates relevance from irrelevance can not map the openness of alternative paths of development, the contingency of alternatives. In causal models, different paths are separately analysed. Each path is connected with chains of consequences which all demand full validity. A network of multiple dependencies can be established. Missing or conflicting assumptions block further causal chains. Those models do not reflect the fact that several alternatives exist or that a factor has to reach a specific value to produce successful alternatives.
When using fuzzy quantifiers like "little" results of the analysis are weak and only conditionally open to verification. Furthermore, it is impossible to combine several statements using fuzzy quantifiers. Consequently, comparing different fuzzy analytical results is difficult. This approach is of no help discovering the relevance of factors.
Both approaches - sharp differentiation between important and unimportant factors as well as using fuzzy quantifiers - provide little help when analysing the impact of different factors on a target criterion. Therefore, it is meaningful to assign numeric values that represent the importance of factors and to find values that represent intermediate levels of relevance between the extremes of "important" and "irrelevant".