This is the first in a series of blog posts analysing and discussing Serbian monthly savings time series.
Aims
The aims of these blog posts are the following:
To describe main features of monthly savings time series
- To model series using popular time series models
- To forecast future savings time series values
- To apply hierarchical time series models
- To apply temporal hierarchical time series models
- To compare forecasting performance of different time series models
Dataset
The dataset under examination comes from the National Bank of Serbia database (http://nbs.rs/export/sites/default/internet/latinica/80/monetarni_sektor/SBMS14.xls). It contains four monthly household savings with banks time series: dinar and FX-indexed savings and foreign currency savings, both in short- and long-terms. All series cover the same period: 2004M1 to 2019M2.
Dataset description
Dataset is available for download in two formats: csv and RData.
Column names in csv file are: Dinar_short, Dinar_long, Foreign_short and Foreign_long. Those familiar with R could download and use RData file. This RData file contains a few time series objects in different format (savings.ts, savings.zoo, sdom, sfor, sshort, and slong). For instance, object sdom contains two series with the following names: “Dinar short-term” and “Dinar long-term”, while object sshort contains two series with the following names: “Dinar short-term” and “Foreign short-term”.
In the next blog post line graphs for each monthly household savings time series will be presented and discussed.
