# Should I be using season as a random or fixed effect?

I am a marine turtle researcher attempting to understand the effect of a harmful algal bloom on our turtle capture rates and the body condition (BC=mass/length^3) of captured turtles.

Field techniques overview: we set a large net (455m long) in the water, in approximately the same spot every time. The amount of time the net remains in the water is recorded, along with the number of captures of two species (loggerhead and green turtle). We conduct these sampling sessions two days per month, year-round.

Harmful algal bloom overview: the HAB consisted of multiple, concurrent events whose effects lasted from April 2011-August 2013.

I am using capture session data from January 2005-December 2015. In order to understand both whether the HAB affected our response variables during the effects of the HAB and whether there are lingering effects after, I have separated the capture session data into three categories: Before (1/2005-3/2011), During (4/2011-8/2013), and After (9/2013-12/2015).

Season (Spring, Summer, Fall, Winter, as defined by solstices and equinoxes) has a strong effect on the number of captures of green turtles. Should I be accounting for this seasonal effect as a random effect or as a fixed effect?

• +1 to both Aksakal comment and mdewey's post. Think of it this way too: Can you have any other season level aside the four you mentioned? Jun 8, 2016 at 18:16