Wilcox test with median 0

[monkeyj]

Hello,
I am trying to run an analysis on differences in female and male aggressive behaviours in an animal species. My data is non-normally distributed and it's both numeric and categorical, so I assumed the Wilcox test will be sufficient. However, all medians come back as 0 and my box and whisker plots come back as flat lines. What does this mean? Have I chosen the wrong statistical test?

 

[hlsmith]

Can you post a histogram for each group for the outcome variable? What is your sample size as well?

 

[monkeyj]

My sample is size is 70. And most data is bound at 0 as I’m measuring the time each animal spent doing a certain action.

 

[hlsmith]

You mean that it is lower bound at 0, or that the animal did not do the task so you gave them a zero?

Also, per your graphs, >50% had a value of zero, correct?

 

[Karabiner]

So the vast majority of indiduals did not engage in aggressive behaviour (zero seconds).
Wouldn't it be much easier and more appropriate to just analyse whether a dicotomous
response variable (aggressive behaviour yes/no) was associated with sex?

With kind regards

Karabiner

 

[monkeyj]

You mean that it is lower bound at 0, or that the animal did not do the task so you gave them a zero?

Also, per your graphs, >50% had a value of zero, correct?

Yes, the majority has the value of 0.

 

[monkeyj]

So the vast majority of indiduals did not engage in aggressive behaviour (zero seconds).
Wouldn't it be much easier and more appropriate to just analyse whether a dicotomous
response variable (aggressive behaviour yes/no) was associated with sex?

With kind regards

Karabiner

That would be most definitely the easiest solution, however, the overall time spent doing the action is unfortunately key.

 

[fed2]

id say the histograms are good enough. maybe add a box-plot.