Data Structure for a Repeated Measures Analysis
March 7, 2000
Repeated-measures designs are very common in many fields of study because of their logistical and statistical efficiency. This newsletter describes repeated-measures designs and the two different ways in which data from a repeated-measures design can be constructed.
A repeated-measures design measures each subject or unit two or more times. For example, one might record weights of children subjected to different diets at several different time points. Or, one might record the heights of a plants to which herbicides have been applied every week for six weeks. In both these examples, two successive measurements on the same child or plant will be expected to be more related to each other than if the measurements were taken from different children or plants. This is the basic reason why a special analysis is required to analyze these data.
Data from a repeated-measures design can be set up in two different ways. In the first way, a row in the data represents a child, referring to the first example. An identification variable is needed for child and another for diet. For each repeated measurement (i.e., weight) over time, a new variable is created. For example:
|
Child
|
Diet
|
Weight1
|
Weight2
|
Weight3
|
|
1
|
1
|
3.4
|
3.9
|
4.0
|
|
2
|
1
|
2.9
|
3.1
|
3.5
|
|
3
|
2
|
3.7
|
3.7
|
4.1
|
In the second way, a row in the data represents a child at a time. Two identification variables are needed for children and times, in addition to an identification variable for diet, but only one variable is needed for the measurements on weight. For example:
|
Child
|
Diet
|
Time
|
Weight
|
|
1
|
1
|
1
|
3.4
|
|
1
|
1
|
2
|
3.9
|
|
1
|
1
|
3
|
4.0
|
|
2
|
1
|
1
|
2.9
|
|
2
|
1
|
2
|
3.1
|
|
2
|
1
|
3
|
3.5
|
|
3
|
2
|
1
|
3.7
|
|
3
|
2
|
2
|
3.7
|
|
3
|
2
|
3
|
4.1
|
Both these methods of structuring the data preserve the connection between the information on the child level and the information on the time level. Which set-up of the data is best in a given situation will depend on a number of factors, including the assumptions to be made, presence of missing data, balance of the design, and software available for analysis.
For further information about how to proceed on constructing or analyzing the data, please contact the Office of Statistical Consulting.
Author: Francoise Vermeylen
Back to StatNews Table of Contents
(This newsletter was distributed to faculty and graduate students in the Division of Nutritional Sciences and the College of Human Ecology, and faculty in the College of Agriculture and Life Sciences, by the Office of Statistical Consulting. Please forward it to any interested colleagues and research staff. Anyone not receiving this newsletter who would like to be added to the mailing list for future newsletters should contact statcons@cornell.edu. Information about the Office of Statistical Consulting can be obtained at World Wide Web address http://www.cscu.cornell.edu.)