Last answered:

24 Feb 2021

Posted on:

29 Dec 2020

0

Time series Analysis for Ages , but not dates

Hello,
How to handle this time series data that has ages rather than dates?
I really want to analyze this time series, but  the data has only ages not dates.  The data shows 'Age' in millions of years and sea level 'Sl' in meters. The present day sea level is the datum zero. Also, the sea level (Sl) have mixed positive & negative values.
How to set the index frequency?
How to use the multiplicative smoothing with negative values?
Thanks in advance

Additional information:

The sea level data in geology is organized & collected as follows. Geologists assume that the present day sea level represents 0 m elevation ( reference point) and the time is zero years as well. So, here we have the present day as a reference/datum in both sea level elevation (SL = 0m) and time (Age = 0 years).

Then we go back to the geological record (studying mostly marine fossils lived at known depths or using stable isotopes) to deduce the sea level elevation back in the geological time ( that goes back to 100s of millions of years). Geologists compare the present day sea level (SL=0 m elevation, Age= 0 time) with the past geological observations in the rock record at very well defined times. If geologists found observations that show the sea level elevations in the past were lower than the present day, they will assign negative sea level elevation values to those ages. On the contrary, if they found the sea level elevation was higher than the present day elevation, they will assig positive sea level elevation values to those ages.

Therefore, Positive/negative sea level elevations are assigned to tell us if the sea level back in the geological time was higher/lower than the present day reference sea level elevation ( SL=0 m elevation, Age= 0 years).

The problem here, all my dates are given as ages( e.g.  1000 years ago, 1000000 years ago) and the sea level elevations are given as negative and positive values relative to the present day reference point ( elevation =0 m, & Age= 0 years).

https://www.youtube.com/watch?v=3XvRe4CYlhQ

https://en.wikipedia.org/wiki/Sea-level_curve

To get the data, please google (haq sea-level curve data)  and this link will show up with the xls file: ( ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/miller2005/miller-sealev-originalcontrib.xls )

Thank you in advance and happy new year! 

======================== Part of The Data below ======
Age Sl
0 0
1000 0
2000 3.3
3000 -4
4000 -4.7
5000 -2
6000 -6.7
7000 -9.3
8000 -12.7
9000 -10
10000 -19.3
11000 -24.7
12000 -46
13000 -55.3
14000 -70
15000 -84
16000 -101.3
17000 -110

2 answers ( 0 marked as helpful)
Instructor
Posted on:

22 Jan 2021

0
Hey Amrouni,    Thanks for reaching out and sorry for the delayed response!   So, for the sea-level metric, I don't think alternating positive and negative values are really an issue if you assume a cyclical trend. Hence, I suggest a Seasonal model will work fine with the data.    As for the dates, you can create another column which contains the corresponding datetime values in reverse. To elaborate, you can take the highest age to be the first position in the time series and today (age = 0) as the max. Then, you can set a frequency of 1000 years to that dataset and set the new column as an index.    I want to help you with this task, so you'll need to give me a hand here. When I load the data using pandas, I get the following columns: 
So, which ones should I use as age here and how are you preprocessing the values? I just want to have my code be as similar to yours before I dive into it.    Best, 365 Vik
Posted on:

24 Feb 2021

0

Yes, Please use only the first column for age and the second column for Sea Level (m).
Thanks you Vik!

Submit an answer