I have a time series of monthly chlorophyll-a observations that spans from 1997 to present day. I have computed the p-value of the time series without removing the seasonality, and it doesn't quite show much of a trend. So, my question is, is the seasonality of the data affecting my computation, and if so, should I remove it? And, any recommendation on which methods that I should use to deseasonalise the time series.
I apologise if my question makes no sense. Thank you in advance.
Well time series include serial correlations, trend & seasonality features. So any prediction or results are conditional on data treatment and/or model. you will need to qualify your conclusions by "de-trended series shows..." or "seasonally adjusted series shows..." or "De-trended and seasonally adjusted series shows...".
***in your question you rightly tell that you fitted linear model and the size of p-value...try models like Fourier curve (single or double)...good luck
Mewa Singh Dhanoa my Achilles' heel is anything that has to do with calculations among many others, but I appreciate your suggestion and shall look into models like Fourier curve.
What do you mean with the p-value of the time series? The p-value is referred to an hypothesis. What is your hypothesis? Or are you talking about
ρ the partial autocorrelation function?
@N. Tartaglione, I’m not certain about the term I am using, sorry. But what, I want to compute is the coefficient of the linear trend. Like in the article, here: 10.1016/j.dsr2.2003.07.018
Trend should be modelled, not removed.
Seasonal coefficients will have their own p-values.
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing