I have a dataset for 'time spent' (hours) carrying out 'behaviours' (walking, lying, standing, grazing), during day or night 'periods' of cattle. Time spent is my dependent variable, with behaviours and periods my independent variables.
Sample dataset below. Time spent is shown as decimal hours and is considered proportional as time spent doing behaviours within any day or night period are dependent on one another (can be expressed as % with sum of all behaviours for a given period totalling 100%). Time spent has a non-normal distribution.
My question is whether or not it is OK to use arcsine (or another) to transform time spent to achieve normally distributed data prior to analysis?
I'm hoping to use a statistical test to determine whether day or night has any affect on time spent carrying out behaviours, but I'm not sure which test I can use.
Cow ID | Period | Behaviour | Hours |
---|---|---|---|
Cow1 | Day | Walking | 4.41 |
Cow1 | Night | Lying | 0 |
Cow1 | Day | Standing | 5.01 |
Cow1 | Night | Grazing | 3.24 |
Cow2 | Day | Walking | 4.41 |
Cow2 | Night | Lying | 2.01 |
Cow2 | Day | Standing | 5.01 |
Cow2 | Night | Grazing | 3.24 |