Page 130 - Quantitative Data Analysis

Quantitative Data Analysis
Simply Explained Using SPSS

3. Range Restriction

Restricting the range of the data limits the variability of data
that can limit the possibility for co-variability between two
variables, thus it will change coefficient of correlation.

The common example of range restriction is Likert scale (1-5).
Range restriction may increase or decrease the correlation.

4. Outlier

Outliers are influential data points that can sharply increase or
decrease correlation. Basically outlier affects the average. Let’s
see, we have 3 observations: 3,5,7, the mean is 5. But if one
point is outlier, say instead of 7 put 27, the means would
increase to 11.67. This sharp increase in mean value affects the
deviation scores, covariance, and variability. Thus the
correlation is highly affected by outlier values. There are certain
methods to detect outliers. However, it is mostly based on
researcher own judgment. If there are few true outliers, then
one or two outlier can change the correlation coefficient.
Therefore, it is highly recommended to analyze and interpret
data with and without outliers. As Pedhazur (1997) said “I
believe that, in addition to
reporting results of analyses
with and without influential
observations, sufficient
information ought to be
given” (p.59). As a last
resort, one can also
transform outlier but it is
not highly recommended.
Four data sets have the same correlation

The Theory and Applications of Statistical Inferences 114