Anova test?

I have data for 09,10, and 2011. I have an exertion level for each individual. My indivduals are classified by the shift they work (8 day, 8 night, 12 day, 12 night). What is the test/method to use to tell if there is a difference between the 4 different shift vars in terms of exertion level (continuous variable). I can plot out a line graph and visualize it, but I it is not clear to me on what analysis to use. The responses are not paired.
Thank you!


No cake for spunky
Are you going to merge the years and compare shifts to exertion levels? Or show seperate results for each year? In that case one way anova (with exertion the dependent variable) will work. If your are going to show trends over time in exertion in a single exercise than I am not sure you can do this with ANOVA. You might need a time series approach.
I understand the anova approach of comparing the means of the exertion levels for the 4 shifts. But I was trying to find a way to NOT merge them but do a single analysis, whatever that may be. The problem is the data that was given to me has no way of being paried throughout the three years. I guess I may have to just do an anova three times (once for each year) and adjust alpha for multiple testing. Do you know any idea what that time series approach may be. I am weak in this area of statistics.


No cake for spunky
I dont understand what you mean by paried.

An alternative to running three anovas (one for each year) is to do a within subject anova with a 2nd factor added. Essentially this would do a within subject ANOVA for the years and seperately (but in the same analysis sort of like a special form of two way anova) look at whether shift influenced the variable. You would also, or could also, see if these interact (that is year and shift).

If you did this you would be assuming essentially that there was some intervention occuring between each year so I don't know if that is a reasonable technqiue for your data. You might ask someone with in depth knowlede of ANOVA about that - I do simple stuff in it :)
By Paired, I mean our data does not have that person "X" said their exertion level was a 8 in 2009, then 7 in 2010 and 6 in 2011. This isn't true longitudinal data. Does this change anything. And thank you for hte suggestion on the within subject anova. Are there any internet references that are helpful on this subject.