Question: Mann-Whitney U test after zero-inflation correction
gravatar for exin
6 months ago by
exin50 wrote:

I'm trying to perform Mann-whitney u test on my data, but realised the excessive zeros cause an issue. I read about the pscl package that uses the zeroinfl function to correct for this, but is unable to find any example online that uses both together. Can I simply perform the test on zeroinfl-corrected data? How do I do that?

Data structure:
time    treat   count
t0  ct  10
t5  ct  1
t10 ct  0
t15 ct  0
t20 ct  0
t25 ct  0
t30 ct  0
t0  ct  7
t5  ct  4
t10 ct  2
t15 ct  2
t20 ct  2
t25 ct  2
t30 ct  2
t0  dm  9
t5  dm  4
t10 dm  0
t15 dm  0
t20 dm  0
t25 dm  0
t30 dm  0
t0  dm  11
t5  dm  10
t10 dm  3
t15 dm  1
t20 dm  1
t25 dm  1
t30 dm  1

wilcox.test(count~treat, data=data, paired=FALSE)
datanew<- zeroinfl(count ~ time+treat, data = data)
R • 228 views
ADD COMMENTlink written 6 months ago by exin50

Let me see if I got it right. There are two replicates from each condition and seven measurement across 30 minutes. IMHO one can't and shouldn't do any statistical test, just plot the four lines and make whatever claim. Any p<0.1 next to it will be worthless or misleading. If there are more replicates, I don't think there is a problem of zero-inflation. Zeros in the data are allowed! It's a ranked test anyway so don't worry about it too much.

ADD REPLYlink modified 6 months ago • written 6 months ago by Asaf6.5k

The statistical test you're using is rank based, it won't care about excess zeros.

ADD REPLYlink written 6 months ago by Devon Ryan92k

Thanks guys!!

Yeh there're more replicates.

ADD REPLYlink written 6 months ago by exin50
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 2054 users visited in the last hour