Thoughts on best method for testing for differences in slopes of lines (time series data)?

Question

I have 4 sets of 50 year data (one value per year for each of the 4 different categories)- and would like to test for differences in the slopes of the lines. I think ANCOVA is appropriate- though maybe not as these are time series data? And ...how to do post hoc pairwise comparisons?

The data are estimates of parasite production for each year under 4 different climate scenarios ranging from hot and dry to cold and wet (parasite production is directly dependent on temperature).

So are your data counts then? – gung - Reinstate Monica Jan 28 '13 at 22:58 — gung - Reinstate Monica, Jan 28 '13 at 22:58

score 1 · Answer 1 · edited Apr 13 '17 at 12:44

1

You are presuming the model form i.e. 1 slope ; no level shift ; no arima structure ; no outliers; no parameter changes over time ; no varaince changes over time. Simply form a robust ARIMA model and estimate it globally and then locally (each of the 4 separately). Conduct a Chow Test as described in my answer to How to compare two time series?

edited Apr 13 '17 at 12:44

Community

1

answered Jan 28 '13 at 22:39

IrishStat

27,906
5
29
55

hmm, thanks! sas even has a proc arima. I'll look into the Chow test, haven't seen that before. – user20203 Jan 28 '13 at 22:41

Thoughts on best method for testing for differences in slopes of lines (time series data)?

1 Answers1