Social Research Glossary A B C D E F G H I J K L M N O P Q R S T U V W X Y Z Home
Citation reference: Harvey, L., 2012-24, Social Research Glossary, Quality Research International, http://www.qualityresearchinternational.com/socialresearch/
This is a dynamic glossary and the author would welcome any e-mail suggestions for additions or amendments.
|
|
_________________________________________________________________
Spuriousness
Spuriousness refers to apparent relationships uncovered by correlation and regression techniques that turn out to be unfounded when other variables, in a multivariate analysis, are taken into account.
Thus an observed relationship between NCO rank and enthusiasm for active military service may be shown to be spurious when length of service is taken into account. In this example, length of service is an antecedent variable that effects both attitude towards battle situations as well as promotion (see below).
A similar situation may arise with an intervening variable. An example, might be the apparent association between sporting events and 'hooliganism' may be explained by the media as an intervening variable that serves to fuel hooligan activity at sporting events.
Spurious, outside the context of multivariate analysis, smply means not what it purports to be, aparently but not actually valid, or fake. For example, the distinction between subjective and objective is spurious.
Example of a spurious relationship
The following example shows how a preceding variable may serve to nullify an observed correlation.
Table 1: Anticipation for combat for a sample of 200 soldiers
High anticipation |
Low anticipation |
Totals |
|
| High rank | 10 |
90 |
100 |
| Low rank | 80 |
20 |
100 |
| Totals | 90 |
110 |
200 |
The hypothetical example (Table 1) is of a sample of 200 hundred armed service draftees and shows their anticipation for combat attitude. There is a high inverse correlation between rank and attitude, such that those of low rank have (surprisingly) a higher (favourable) anticipation for combat. However, if a control variable 'length of service' is introduced, the table is split as follows (Table 2 and Table 3):
Table 2: Anticipation for combat for the sub-sample (n=100) of the 200 soldiers with less than two years service
High anticipation |
Low anticipation |
Totals |
|
| High rank | 12 |
8 |
20 |
| Low rank | 48 |
32 |
80 |
| Totals | 60 |
40 |
100 |
Table 3: Anticipation for combat for the sub-sample (n=100) of the 200 soldiers with more than two years service
High anticipation |
Low anticipation |
Totals |
|
| High rank | 24 |
56 |
80 |
| Low rank | 6 |
14 |
20 |
| Totals | 30 |
70 |
100 |
Once the sample has been split into two groups based on the control variable 'length of service', one may note that the correlation disappears, sixty per cent of those with less than two years service had high anticipation irrespective of rank (Table 2). Similarly only 30 per cent of those with more than two years service had high anticipation, irrespective of rank (Table 3). The association between rank and anticipation is explained away by the prior control variable 'length of service'.
See also
Researching the Real World Section 8
Top