Chapter 11 – Relationships Can Be Deceiving
Correlation does not provide the complete picture. Many things can cause correlation. Others can
inflate and deflate correlations. Groups combined inappropriately can mask relationships. Small samples
are highly affected by outliers. Consistent outliers inflate the correlation, inconsistent outliers deflate it.
Researchers usually check for initially recorded outliers. Outliers can occur as legitimate data. Some
researchers may leave them in to lead you astray. Illegitimate correlation can occur when two or more
groups are combined when they should be considered separately. Example: hardcover and soft cover
books in price/page study. Together it is negative correlation, separately they are positive correlation.
Correlation does not imply causation. Relationships and correlations from observed studies are reported
as causal. Sometimes connections do make sense, but if there is no evidence we can’t prove causation.
Some reasons that variables can be related include:
The explanatory variable is a direct cause of the response variable. Even though they are related there
may not be a strong association between the two variables.
The response variable is causing a change in the explanatory variable. Causal connections can be
opposite. Example: decrease in occupancy in a hotel means an increase in advertising.