non-parametric t-test with mean/SD = 0

#1
Hi All, quick question on non-parametric t-test

I have 2 groups. 7 values each.

Group One= 0, 0, 0, 0, 0, 0, 0 which of course is mean=0 and SEM=0

Group Two= 0, 0, 21, 35, 0, 0, 0 mean=8 and SEM=5.4

When I do unpaired non-parametric Mann-Whitney two-tail test, I am told that because the SEM/SD value is 0 in Group One, the data cannot be analyzed.

My question is how to by pass this? Is the data significantly different anyway since the mean of Group One is 0?

Thanks in advance
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
What program are you using? Why aren't you comparing medians within this test? For other analyses, people sometimes add a constant to all values to get past similar errors. This can be referred to as a continuity correction, an example may be adding 0.5 to all values in contingency table data.
 

Dason

Ambassador to the humans
#4
Can you give some more information about your data? Where does it come from? It seems to be count data - how was it collected?
 
#5
Thanks for everyones replies. I am using Instat 3 from GraphPad 5

The data is bacterial count from animal organs. I tried adding a small value like 0.01 or 1 Then the average and SEM does change. But the post t-test says the data is not significant. Because I am doing the same Mann-whitney test for all other groups and data, I did not want to change the values to median or do a wilcoxon test (that would make the manuscript more complicated).
 
#6
Im not sure how to do a one-sample Wilcoxon on Instat 3. I tried with wilcoxon matched paired test (adding a '1' value to my first group so my mean would and SEM would not be '0'), but it says not significant. I guess the big problem is that my second group has too many 0 values giving it a large SD/SEM

A way is to run a one-sample Wilcoxon test. In that case you will compare the second group with the value zero.