The blue moving average line is mostly responding to the peak at the end of 2016. The last 3 points going up is mostly noise. I would smooth over a longer period because the small-scale variability is autocorrelated and then the trend would still be down.
I do not know whether the correlations length is that long, but just looks at the deviations from the blue "trend" line: if one week is high, the next week is high and visa versa. Just by eye I would say that that goes for several months. Then averaging over just one year, means that you just have a few samples and the "trend" is not supposed to be noisy.
The blue moving average line is mostly responding to the peak at the end of 2016. The last 3 points going up is mostly noise. I would smooth over a longer period because the small-scale variability is autocorrelated and then the trend would still be down.
ReplyDeleteVictor: Thanks for your comment.
ReplyDeleteBut how is coal production data autocorrelated?
How does what they produce in the last week depend on what they produced in the week a year ago?
I do not know whether the correlations length is that long, but just looks at the deviations from the blue "trend" line: if one week is high, the next week is high and visa versa. Just by eye I would say that that goes for several months. Then averaging over just one year, means that you just have a few samples and the "trend" is not supposed to be noisy.
ReplyDelete