A ceiling or floor describes an uneven distribution at the high or low end of a range (respectively). There is a clustering of values with none above (for a ceiling) or none below (for a floor).


Significance of a ceiling or floor:

(1) The method does not discriminate between elements within the cluster. This may occur for a number of reasons, such as the test was too hard (for a floor) or too easy (for a ceiling).

(2) Analyzing the results with a method that assumes a normal distribution or a linear response will give misleading results.

(3) A change in the underlying object of interest may not be reflected by a change in its score.


Ways to approach a method showing a ceiling or floor effect:

(1) Redesign the method so that there is discrimination of the clustering elements.

(2) Select an appropriate method for analyzing the data.

(3) Do not use the original method as a means of following change over time.


To read more or access our algorithms and calculators, please log in or register.