Christoph Lehmann
2003-Aug-19 08:15 UTC
[R] test for equal distributions with small numbes of observations
Dear R-pros I have a problem, for which usually I would apply a chisq.test or a fisher.test: 40 objects, each given either a "0" or "1", regarding if this object later on will be remembered by a subject or not. 7 subjects investigated means: we have a 2x40 matrix, each cell the number of subjects for who the object i has been given either "0" or "1" e.g. objects: 1 1 3 39 40 ------------------------------------------ "0" 1 2 2 .. 7 7 "1" 4 4 5 .. 0 0 over all 40 objects, we have 67% of "1" and 33% of "0" I want to know, if for the 40 objects, the ratio of "0"/"1" differs or not, i.e. if they have the same distribution. I cannot use a chisq.test since the expected frequencies are < 5 for the "0" cells. Fisher.test seems to run for > 12h on a PIV 1.8GHz... what do you recommend me to do? Many thanks Christoph --> recognition[,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10] [,11] [,12] [,13] [,14] [1,] 1 2 2 2 2 3 3 3 3 3 3 3 4 4 [2,] 4 4 5 4 4 3 4 4 4 4 3 4 3 3 [,15] [,16] [,17] [,18] [,19] [,20] [,21] [,22] [,23] [,24] [,25] [,26] [1,] 4 4 4 4 4 4 5 5 5 5 5 5 [2,] 3 2 3 3 2 3 2 2 2 2 2 2 [,27] [,28] [,29] [,30] [,31] [,32] [,33] [,34] [,35] [,36] [,37] [,38] [1,] 5 5 5 5 5 6 6 6 6 6 6 6 [2,] 2 2 2 2 2 1 1 1 1 1 1 1 [,39] [,40] [1,] 7 7 [2,] 0 0> fisher.test(recognition)-- Christoph Lehmann <christoph.lehmann at gmx.ch>
Carlos J. Gil Bellosta
2003-Aug-19 09:04 UTC
[R] test for equal distributions with small numbes of observations
Probably, to check homogenity you only need to consider a vector of dimension 8 consisting in the number of times you get any of the configurations (0,7), (1,6), (2,5),...(7,0). This is most likely a sufficient statistics. Under homogeneity, it should be distributed according to a multinomial random variable with probability vector equal to the density of a binomial variable of unknown parameter p. Then you can use standard tests to see if the MVE fits or if it does not fit the data. Carlos J. Gil Bellosta Sigma Consultores Estad?sticos http://www.consultoresestadisticos.com Christoph Lehmann wrote:>Dear R-pros > >I have a problem, for which usually I would apply a chisq.test or a >fisher.test: > > > >40 objects, each given either a "0" or "1", regarding if this object >later on will be remembered by a subject or not. > >7 subjects investigated > >means: we have a 2x40 matrix, each cell the number of subjects for who >the object i has been given either "0" or "1" e.g. > >objects: > 1 1 3 39 40 > ------------------------------------------ >"0" 1 2 2 .. 7 7 >"1" 4 4 5 .. 0 0 > >over all 40 objects, we have 67% of "1" and 33% of "0" > >I want to know, if for the 40 objects, the ratio of "0"/"1" differs or >not, i.e. if they have the same distribution. > >I cannot use a chisq.test since the expected frequencies are < 5 for the >"0" cells. > >Fisher.test seems to run for > 12h on a PIV 1.8GHz... > >what do you recommend me to do? > >Many thanks > >Christoph >-- > > >>recognition >> >> > [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10] [,11] [,12] [,13] [,14] >[1,] 1 2 2 2 2 3 3 3 3 3 3 3 4 4 >[2,] 4 4 5 4 4 3 4 4 4 4 3 4 3 3 > [,15] [,16] [,17] [,18] [,19] [,20] [,21] [,22] [,23] [,24] [,25] [,26] >[1,] 4 4 4 4 4 4 5 5 5 5 5 5 >[2,] 3 2 3 3 2 3 2 2 2 2 2 2 > [,27] [,28] [,29] [,30] [,31] [,32] [,33] [,34] [,35] [,36] [,37] [,38] >[1,] 5 5 5 5 5 6 6 6 6 6 6 6 >[2,] 2 2 2 2 2 1 1 1 1 1 1 1 > [,39] [,40] >[1,] 7 7 >[2,] 0 0 > > > >>fisher.test(recognition) >> >> > > > >
kjetil brinchmann halvorsen
2003-Aug-20 02:49 UTC
[R] test for equal distributions with small numbes of observations
On 19 Aug 2003 at 10:15, Christoph Lehmann wrote:> Dear R-pros > > I have a problem, for which usually I would apply a chisq.test or a > fisher.test: >try ?chisq.test which tells you you can say chisq.test(..., sim=TRUE, B=20000) and the simulation will be very fast! Kjetil Halvorsen> > > 40 objects, each given either a "0" or "1", regarding if this object > later on will be remembered by a subject or not. > > 7 subjects investigated > > means: we have a 2x40 matrix, each cell the number of subjects for who > the object i has been given either "0" or "1" e.g. > > objects: > 1 1 3 39 40 > ------------------------------------------ > "0" 1 2 2 .. 7 7 > "1" 4 4 5 .. 0 0 > > over all 40 objects, we have 67% of "1" and 33% of "0" > > I want to know, if for the 40 objects, the ratio of "0"/"1" differs or > not, i.e. if they have the same distribution. > > I cannot use a chisq.test since the expected frequencies are < 5 for the > "0" cells. > > Fisher.test seems to run for > 12h on a PIV 1.8GHz... > > what do you recommend me to do? > > Many thanks > > Christoph > -- > > recognition > [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10] [,11] [,12] [,13] [,14] > [1,] 1 2 2 2 2 3 3 3 3 3 3 3 4 4 > [2,] 4 4 5 4 4 3 4 4 4 4 3 4 3 3 > [,15] [,16] [,17] [,18] [,19] [,20] [,21] [,22] [,23] [,24] [,25] [,26] > [1,] 4 4 4 4 4 4 5 5 5 5 5 5 > [2,] 3 2 3 3 2 3 2 2 2 2 2 2 > [,27] [,28] [,29] [,30] [,31] [,32] [,33] [,34] [,35] [,36] [,37] [,38] > [1,] 5 5 5 5 5 6 6 6 6 6 6 6 > [2,] 2 2 2 2 2 1 1 1 1 1 1 1 > [,39] [,40] > [1,] 7 7 > [2,] 0 0 > > > fisher.test(recognition) > > > -- > Christoph Lehmann <christoph.lehmann at gmx.ch> > > ______________________________________________ > R-help at stat.math.ethz.ch mailing list > https://www.stat.math.ethz.ch/mailman/listinfo/r-help