Score test for zero inflation in Poisson data
Carries out a simple score test (van den Broek, 1995) for excess zeros in an otherwise Poisson distribution of counts. It gives a χ^2_1 statistic on one degree of freedom.
zero.test(x)
x |
A vector of non-negative counts, or a one-way frequency table of such counts. |
The test first calculates the rate estimate from the mean, \hat{λ} = \bar{x}. The number of observed zeros, n_0 is then compared with the expected number, n \hat{p_0}, where \hat{p}_0=\exp[-\hat{λ}]. Then the test statistic is calculated by the formula:
\frac{(n_0 - n\hat{p}_0)^2}{n\hat{p}_0(1-\hat{p}_0) - n\bar{x}\hat{p}_0^2}
This test statistic has a χ^2_1 distribution.
Returns invisibly a list of three elements:
|
Description of 'comp1' |
|
Description of 'comp2' |
|
Upper tail p-value |
Michael Friendly
The original R code came from a Stackexchange question, https://stats.stackexchange.com/questions/118322/how-to-test-for-zero-inflation-in-a-dataset
Van den Broek, J. (1995). A Score Test for Zero Inflation in a Poisson Distribution. Biometrics, 51(2), 738-743. doi: 10.2307/2532959
Yang, Zhao, James W. Hardin, and Cheryl L. Addy (2010). Score Tests for Zero-Inflation in Overdispersed Count Data. Communications in Statistics - Theory and Methods 39 (11) 2008-2030. doi: 10.1080/03610920902948228
# synthetic tests zero.test(rpois(100, 1)) zero.test(rpois(100, 5)) # add some extra zeros zero.test(c(rep(0, 20), rpois(100, 5))) # Articles by Phd candidates data(PhdPubs, package="vcdExtra") zero.test(PhdPubs$articles) phd.tab <- table(PhdPubs$articles) zero.test(phd.tab)
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.