Which nonparametric test to use for comparing two related samples?

#1
Hi,
I am looking to compare day and night time depths for basking sharks in SPSS. I found that if I have day depths for all sharks together I can use a Mann Whitney U test as then the data is independent (they are non normally distributed). However, I also need to compare each shark individually to see if there is a significant difference between depth for day and night. I was thinking of using a Wilcoxon Signed Ranks test as since this data is from one individual then the samples must be dependent. However my data is not paired exactly I just have a list of depths for day and a list for night. I was just wondering is this the right test to use or could I still use a Mann Whitney U test?
Thanks.
 
#2
However, I also need to compare each shark individually to see if there is a significant difference between depth for day and night. I was thinking of using a Wilcoxon Signed Ranks test as since this data is from one individual then the samples must be dependent. However my data is not paired exactly I just have a list of depths for day and a list for night.
I couldn't understand what is going on exactly. Maybe more clarifications can help. But anyhow I think a Mann-Whitney is still OK.
 
#3
I also need to compare each shark individually to see if there is a significant difference between depth for day and night. I was thinking of using a Wilcoxon Signed Ranks test as since this data is from one individual then the samples must be dependent. However my data is not paired exactly I just have a list of depths for day and a list for night. I was just wondering is this the right test to use or could I still use a Mann Whitney U test?
Thanks.
If each shark has more than one measurement for day and night, then you can use Wilcoxon Signed Rank test (the paired test version of Wilcoxon Rank Sum/MWU) since your individual measurements would be paired. Depending on how many measurements you have, consider using continuity correction if you only have a few measurements. Also keep in mind that t-test is remarkably robust, even for small sample size data that slightly deviate from normality.