Spurious relationship

Hypothesis testing

Often one tests a null hypothesis of no correlation between two variables, and chooses in advance to reject the hypothesis if the correlation computed from a data sample would have occurred in less than (say) 5% of data samples if the null hypothesis were true. While a true null hypothesis will be accepted 95% of the time, the other 5% of the times having a true null of no correlation a zero correlation will be wrongly rejected, causing acceptance of a correlation which is spurious (an event known as Type I error). Here the spurious correlation in the sample resulted from random selection of a sample that did not reflect the true properties of the underlying population.

Detecting spurious relationships

The term "spurious relationship" is commonly used in statistics and in particular in experimental research techniques, both of which attempt to understand and predict direct causal relationships (X → Y). A non-causal correlation can be spuriously created by an antecedent which causes both (W → X and W → Y). Mediating variables, (X → M → Y), if undetected, estimate a total effect rather than direct effect without adjustment for the mediating variable M. Because of this, experimentally identified correlations do not represent causal relationships unless spurious relationships can be ruled out.

Experiments

In experiments, spurious relationships can often be identified by controlling for other factors, including those that have been theoretically identified as possible confounding factors. For example, consider a researcher trying to determine whether a new drug kills bacteria; when the researcher applies the drug to a bacterial culture, the bacteria die. But to help in ruling out the presence of a confounding variable, another culture is subjected to conditions that are as nearly identical as possible to those facing the first-mentioned culture, but the second culture is not subjected to the drug. If there is an unseen confounding factor in those conditions, this control culture will die as well, so that no conclusion of efficacy of the drug can be drawn from the results of the first culture. On the other hand, if the control culture does not die, then the researcher cannot reject the hypothesis that the drug is efficacious.

Non-experimental statistical analyses

Disciplines whose data are mostly non-experimental, such as economics, usually employ observational data to establish causal relationships. The body of statistical techniques used in economics is called econometrics. The main statistical method in econometrics is multivariable regression analysis. Typically a linear relationship such as

${\displaystyle y=a_0+a_1{x}_1+a_2{x}_2+\cdots +a_k{x}_k+e}$

is hypothesized, in which ${\displaystyle y}$ is the dependent variable (hypothesized to be the caused variable), ${\displaystyle x_j}$ for j = 1, ..., k is the j^th independent variable (hypothesized to be a causative variable), and ${\displaystyle e}$ is the error term (containing the combined effects of all other causative variables, which must be uncorrelated with the included independent variables). If there is reason to believe that none of the ${\displaystyle x_j}$s is caused by y, then estimates of the coefficients ${\displaystyle a_j}$ are obtained. If the null hypothesis that ${\displaystyle a_j=0}$ is rejected, then the alternative hypothesis that ${\displaystyle a_j\ne 0}$ and equivalently that ${\displaystyle x_j}$ causes y cannot be rejected. On the other hand, if the null hypothesis that ${\displaystyle a_j=0}$ cannot be rejected, then equivalently the hypothesis of no causal effect of ${\displaystyle x_j}$ on y cannot be rejected. Here the notion of causality is one of contributory causality: If the true value ${\displaystyle a_j\ne 0}$, then a change in ${\displaystyle x_j}$ will result in a change in y unless some other causative variable(s), either included in the regression or implicit in the error term, change in such a way as to exactly offset its effect; thus a change in ${\displaystyle x_j}$ is not sufficient to change y. Likewise, a change in ${\displaystyle x_j}$ is not necessary to change y, because a change in y could be caused by something implicit in the error term (or by some other causative explanatory variable included in the model).

Regression analysis controls for other relevant variables by including them as regressors (explanatory variables). This helps to avoid mistaken inference of causality due to the presence of a third, underlying, variable that influences both the potentially causative variable and the potentially caused variable: its effect on the potentially caused variable is captured by directly including it in the regression, so that effect will not be picked up as a spurious effect of the potentially causative variable of interest. In addition, the use of multivariate regression helps to avoid wrongly inferring that an indirect effect of, say x₁ (e.g., x₁ → x₂ → y) is a direct effect (x₁ → y).

Just as an experimenter must be careful to employ an experimental design that controls for every confounding factor, so also must the user of multiple regression be careful to control for all confounding factors by including them among the regressors. If a confounding factor is omitted from the regression, its effect is captured in the error term by default, and if the resulting error term is correlated with one (or more) of the included regressors, then the estimated regression may be biased or inconsistent (see omitted variable bias).

In addition to regression analysis, the data can be examined to determine if Granger causality exists. The presence of Granger causality indicates both that x precedes y, and that x contains unique information about y.

Other relationships

There are several other relationships defined in statistical analysis as follows.

• Direct relationship
• Mediating relationship
• Moderating relationship

Literature

• David A. Freedman (1983) A Note on Screening Regression Equations, The American Statistician, 37:2, 152–155, DOI: 10.1080/00031305.1983.10482729