Ederer-Myers-Mantel test
|
|
|
|
Indications/Recommendations for use: When cases are clustered
m1 will be large, m1 will be small
when cases occur uniformly through time. Either of these alternatives
(m1 larger than expected or m1
smaller than expected) will cause the statistic to be large (more).
|
|
Description: A test for time clustering in several time series
simultaneously.
|
|
Test statistic: The test statistic is m1, the maximum number of cases in a time series. When cases are clustered m1 will be large; it will be small when cases occur uniformly through time. The data from several time series (each time series represents an area such as a county) are used to construct a Chi-square statistic to test for time clustering in several areas simultaneously:
|
|
|
Null Hypothesis: Cases occur at random in each time series.
|
|
Alternative Hypothesis:
Cases do not occur randomly through time, they either cluster or occur
uniformly.
|
|
GeoMed Inputs:
Counts of cases in consecutive time periods (a time series) for several areas. An
example is the number of cases of leukemia in 8 counties by month over 1 year. Notes: The test is
insensitive to different population sizes over the areas, but is biased by changes in population size
through time.
The Ederer - Myers - Mantel test requires the name of a time series (TIM) file.
|
GeoMed Outputs:
- r, the total number of cases in a time series
- f(r), the number of time series with that number of cases
- E(m1) and Var(m1) under the null hypothesis for a time series of similar length and r
- Chi square value and significance level
- Plot of expected m1 on the observed m1
The mean and variance of the expected number of disease cases in each
of the possible area-time interval cells is used to evaluate the
significance of the EMM test.
|
|
Example Analysis
|
Reference:
Stark, C. R. and N. Mantel. 1967. Lack of seasonal or temporal
spatial clustering of Down's Syndrome births in Michigan. American
Journal of Epidemiology 86: 199-213.
|
