Database Reference
In-Depth Information
terminology
What it meanS
Correlation and
causation
Even though there may be a relationship between two
things (correlation), there isn't evidence that one thing
directly causes the other thing to happen (causation).
Suntan lotion sales and ice cream sales both go up at the
same time. It doesn't mean that suntan lotion causes you
to crave ice cream. They're both related to temperature,
the common factor driving sales of both.
Statistical significance
You want to know if you can be certain that the results
are not due to chance—that they really are a result of an
intervention, action, or explanation. Statistical significance
means that a finding or value is reliable. That is, the result
isn't likely explained by random chance.
Over-indexing,
under-indexing
It is often helpful to view values through the lens of an
index, where the index may be an average, benchmark, or
standard value. In some cases, the index is set to 100, so
it is easy to compare an over-indexing value (greater than
100) or under-indexing values (less than 100).
For example, a Pew survey showed that 26 percent of
African-American Internet users surveyed said they used
Twitter, compared to 14 percent of Caucasian users and
19 percent of Hispanics. 1 In this example, African-American
users over-index on Twitter.
Regression model
A statistical approach to determine the relationship
between variables. Typically many independent vari-
ables are included in the regression model in an effort to
explain movements in a dependent variable. For example,
many descriptive factors about customers may be used
in a model to try to predict the likelihood of a customer
leaving.
Generalizable
A determination of whether the results from the study/
survey may be applied to a broader population than was
included in the group who actually participated in the
study/survey.
Regression to
the mean
Are the results influenced by the tendency for things to
move back toward the average? For example, initial results
may be impacted by a unique performance instead of a
person's average performance over time, and what we are
now seeing is a person's typical performance.
Confounding factors
A situation in which a variable that helps explain a rela-
tionship has been left out of a statistical model. For exam-
ple, a perceived relationship between two factors (suntan
location sales and ice cream sales) may be misunderstood
for lack of including a confounding factor (temperature).
Search WWH ::




Custom Search