Spurious relationships

Bennett (1994) argues that poor people are more subject to disease and more likely to live near electric-power eyesores, such as substations. Thus, we find a spurious relationship between proximity to electric-power facilities and cancer. 

Internal validity Seeking to establish a causal relationship, whereby certain conditions are believed to lead to other conditions, as distinguished from spurious relationships (for explanatory or causal studies only and not for descriptiOTi or exploratory study). 

It is rather amazing that we have only recently learnt how to use this abundance of data. Traditional statistics has wrongly taught us that we must always have fewer variables (K) than observations (N), unless we should fall into a morass of spurious relationships. This caution is correct if we are to treat each variable as being precise and independent (i.e., having some unique piece of information). 

My second example is just as egregious in its own way. White (1982), in what is otherwise an exemplary and still valuable meta-analysis, discusses the use of SES both as a spurious influence (covariate or nuisance variable) and as a causal agent (p. 462). He completely fails to warn the reader, however, that it is mandatory to sort correlations into those based on relationships between biological relatives and those based on relationships between adopted parents and their adopted children. His failure to discuss such a critical issue is perplexing as he does cite two classic adoption studies (Burks, 1928a Freeman, Holzinger, Mitchell, 1928) but not that of Scarr and Weinberg published in 1978. 

As discussed in previous paragraphs, the spectral absorption of the chromophores of vision is much more comphcated than that described by Beer s Law for true solutions of low molarity. The absorption of the chromophores of vision is a function of the environment, the chemical state, the spatial relationship, and the orientation of the molecules. They actually exhibit additional, and generally dominant, absorption spectra in the liquid crystalline state that are not found for the same material in low molarity solution. These additional spectra are highly anisotropic. They also exhibit a high absorption coefficient along the preferred axis. Because of these properties, it is extremely difficult to make a comparison of the properties of the chromophores when they are in-vivo with their properties when in-vitro. Because the chromophores are not in solution, when in-vivo, it is not appropriate to use Lambert-Beer s Law to evaluate them (See Section 5.3.5.3), The results of using this law are generally spurious. 

Davis et al. (2000) discuss the problem of correcting certain pharmacokinetic values for the fraction unbound in plasma, advising caution in the approach as it may lead to spurious correlations. Values for clearance or volume of distribution of compounds have been reported in previous studies to correlate with log K. In some of these studies the correlations use values for clearance or volume of distribution that have been corrected for the fraction unbound in plasma. However, the fraction unbound in plasma is itself dependent on log K. Therefore, it is important to check that these reported correlations are genuine and not merely a reflection of the relationship between plasma protein binding and log K. ... 

Supposed isokinetic relationships as established by direct correlation of Awith A S are often spurious, and the calculated value of P is meaningless, because errors in Alead to compensating errors in A S. Satisfactory methods of establishing such relationships have been devised. 

---