[](https://git.dynare.org/Dynare/dseries/commits/master)
This MATLAB/Octave toolbox comes with two classes:
- `@dates` which is used to handle dates.
- `@dseries` which is used to handle time series data.
The package is a dependence of
[Dynare](https=//git.dynare.org/Dynare/dynare), but can also be used
as a standalone package without Dynare. The package is
compatible with MATLAB 2008a and following versions, and (almost
compatible with) the latest Octave version.
## Installation
The toolbox can be installed by cloning the Git repository:
~$ git clone https://git.dynare.org/Dynare/dseries.git
or downloading a zip archive:
~$ wget https://git.dynare.org/Dynare/dseries/-/archive/master/dseries-master.zip
~$ unsip dseries-master.zip
-$ mv dseries-master dseries
## Usage
Add the `dseries/src` folder to the MATLAB/Octave path, and run the following command (on MATLAB/Octave) prompt:
>> dseries().initialize()
which, depending on your system, will add the necessary subfolders to
the MATLAB/Octave path.
You are then ready to go. A full documentation will come soon,
but you can already obtain a general idea by looking into the Dynare
reference manual.
Note that [X13-ARIMA-SEATS](https://www.census.gov/srd/www/x13as/) is required
for accessing all the features of the toolbox. On Windows and macOS, an
X13-ARIMA-SEATS binary is included in standalone dseries packages and in Dynare
packages. On Debian and Ubuntu it is possible to install X13-ARIMA-SEATS with
`apt install x13as` (on Debian, you must have the non-free archive area listed
in package sources).
## Examples
### Instantiate a dseries object from an array
>> A = randn(50, 3);
>> d = dseries(A, dates('2000Q1'), {'A1', 'A2', 'A3'});
The first argument of the `dseries` constructor is an array of data,
observations and variables are respectively along the rows and
columns. The second argument is the initial period of the dataset. The
last argument is a cell array of row character arrays for the names of
the variables.
>> d
d is a dseries object:
| A1 | A2 | A3
2000Q1 | -1.0891 | -2.1384 | -0.29375
2000Q2 | 0.032557 | -0.83959 | -0.84793
2000Q3 | 0.55253 | 1.3546 | -1.1201
2000Q4 | 1.1006 | -1.0722 | 2.526
2001Q1 | 1.5442 | 0.96095 | 1.6555
2001Q2 | 0.085931 | 0.12405 | 0.30754
2001Q3 | -1.4916 | 1.4367 | -1.2571
2001Q4 | -0.7423 | -1.9609 | -0.86547
2002Q1 | -1.0616 | -0.1977 | -0.17653
2002Q2 | 2.3505 | -1.2078 | 0.79142
| | |
2009Q4 | -1.7947 | 0.96423 | 0.62519
2010Q1 | 0.84038 | 0.52006 | 0.18323
2010Q2 | -0.88803 | -0.020028 | -1.0298
2010Q3 | 0.10009 | -0.034771 | 0.94922
2010Q4 | -0.54453 | -0.79816 | 0.30706
2011Q1 | 0.30352 | 1.0187 | 0.13517
2011Q2 | -0.60033 | -0.13322 | 0.51525
2011Q3 | 0.48997 | -0.71453 | 0.26141
2011Q4 | 0.73936 | 1.3514 | -0.94149
2012Q1 | 1.7119 | -0.22477 | -0.16234
2012Q2 | -0.19412 | -0.58903 | -0.14605
>>
### Instantiate a dseries object from a file
It is possible to instantiate a `dseries` object from a `.csv`,
`.xls`, `.xlsx`, `.mat` or `m` file, see the Dynare reference manual
for a complete description of the constraints on the content of these
files.
>> websave('US_CMR_data_t.csv', 'https://www.dynare.org/Datasets/US_CMR_data_t.csv');
>> d = dseries('US_CMR_data_t.csv');
>> d
d is a dseries object:
| gdp_rpc | conso_rpc | inves_rpc | defgdp | ... | networth_rpc | re | slope | creditspread
1980Q1 | 47941413.1257 | NaN | NaN | 0.40801 | ... | 33.6814 | 0.15047 | -0.0306 | 0.014933
1980Q2 | 46775570.3923 | NaN | NaN | 0.41772 | ... | 32.2721 | 0.12687 | -0.0221 | 0.028833
1980Q3 | 46528261.9561 | NaN | NaN | 0.42705 | ... | 36.6499 | 0.098367 | 0.011167 | 0.022167
1980Q4 | 47249592.2997 | NaN | NaN | 0.43818 | ... | 39.4069 | 0.15853 | -0.0343 | 0.022467
1981Q1 | 48059176.868 | NaN | NaN | 0.44972 | ... | 37.9954 | 0.1657 | -0.0361 | 0.0229
1981Q2 | 47531422.174 | NaN | NaN | 0.45863 | ... | 38.6262 | 0.1778 | -0.0403 | 0.0202
1981Q3 | 47951509.5055 | NaN | NaN | 0.46726 | ... | 36.3246 | 0.17577 | -0.0273 | 0.016333
1981Q4 | 47273009.6902 | NaN | NaN | 0.47534 | ... | 34.8693 | 0.13587 | 0.005 | 0.025933
1982Q1 | 46501690.1111 | NaN | NaN | 0.48188 | ... | 32.0964 | 0.14227 | 0.00066667 | 0.027367
1982Q2 | 46525455.3206 | NaN | NaN | 0.48814 | ... | 31.6967 | 0.14513 | -0.0058333 | 0.0285
| | | | | ... | | | |
2016Q1 | 85297205.4011 | 51926452.5716 | 21892729.0934 | 1.0514 | ... | 420.7154 | 0.0016 | 0.0203 | 0.0323
2016Q2 | 85407205.5913 | 52096454.9154 | 21824323.7487 | 1.0506 | ... | 398.7084 | 0.0036 | 0.0156 | 0.0339
2016Q3 | 85796604.1157 | 52436447.9843 | 21874814.014 | 1.0578 | ... | 424.8703 | 0.0037333 | 0.0138 | 0.029167
2016Q4 | 86101149.6919 | 52595613.0404 | 22010921.8985 | 1.0617 | ... | 444.622 | 0.0039667 | 0.011667 | 0.026967
2017Q1 | 86376652.4732 | 52795431.0988 | 22399301.0801 | 1.0672 | ... | 450.8777 | 0.0045 | 0.0168 | 0.0251
2017Q2 | 86982016.8089 | 53164725.076 | 22671020.5449 | 1.0728 | ... | 481.8778 | 0.007 | 0.017433 | 0.022167
2017Q3 | 87605975.0339 | 53451779.0342 | 23033324.7981 | 1.0758 | ... | 496.3342 | 0.0095 | 0.013133 | 0.022367
2017Q4 | 88111231.6601 | 53601437.7291 | 23477516.6946 | 1.081 | ... | 509.1968 | 0.011533 | 0.0109 | 0.020867
2018Q1 | 88557263.9759 | 53960814.0875 | 23726936.444 | 1.0882 | ... | 536.4746 | 0.012033 | 0.011667 | 0.019
2018Q2 | 88817646.3122 | 53931032.9449 | 23989494.0402 | 1.0937 | ... | 560.3093 | 0.014467 | 0.013133 | 0.0171
2018Q3 | 89689102.8539 | 54343965.1391 | 24123408.6269 | 1.1027 | ... | 554.472 | 0.017367 | 0.011833 | 0.0186
>>
### Create time series
Using an existing `dseries` object it is possible to create new time series:
>> d.cy = d.conso_rpc/d.gdp_rpc
d is a dseries object:
| conso_rpc | creditspread | cy | defgdp | ... | pinves_defl | re | slope | wage_rph
1980Q1 | NaN | 0.014933 | NaN | 0.40801 | ... | 145.6631 | 0.15047 | -0.0306 | 65.0376
1980Q2 | NaN | 0.028833 | NaN | 0.41772 | ... | 145.6095 | 0.12687 | -0.0221 | 65.1872
1980Q3 | NaN | 0.022167 | NaN | 0.42705 | ... | 145.3811 | 0.098367 | 0.011167 | 65.3858
1980Q4 | NaN | 0.022467 | NaN | 0.43818 | ... | 144.3745 | 0.15853 | -0.0343 | 65.5028
1981Q1 | NaN | 0.0229 | NaN | 0.44972 | ... | 144.6055 | 0.1657 | -0.0361 | 65.4385
1981Q2 | NaN | 0.0202 | NaN | 0.45863 | ... | 145.6512 | 0.1778 | -0.0403 | 65.3054
1981Q3 | NaN | 0.016333 | NaN | 0.46726 | ... | 144.7545 | 0.17577 | -0.0273 | 65.5074
1981Q4 | NaN | 0.025933 | NaN | 0.47534 | ... | 145.4748 | 0.13587 | 0.005 | 65.4142
1982Q1 | NaN | 0.027367 | NaN | 0.48188 | ... | 144.924 | 0.14227 | 0.00066667 | 66.1617
1982Q2 | NaN | 0.0285 | NaN | 0.48814 | ... | 144.4647 | 0.14513 | -0.0058333 | 65.8827
| | | | | ... | | | |
2016Q1 | 51926452.5716 | 0.0323 | 0.60877 | 1.0514 | ... | 98.7988 | 0.0016 | 0.0203 | 102.4176
2016Q2 | 52096454.9154 | 0.0339 | 0.60998 | 1.0506 | ... | 98.2923 | 0.0036 | 0.0156 | 102.5282
2016Q3 | 52436447.9843 | 0.029167 | 0.61117 | 1.0578 | ... | 98.1811 | 0.0037333 | 0.0138 | 102.0061
2016Q4 | 52595613.0404 | 0.026967 | 0.61086 | 1.0617 | ... | 98.0833 | 0.0039667 | 0.011667 | 102.1861
2017Q1 | 52795431.0988 | 0.0251 | 0.61122 | 1.0672 | ... | 97.8223 | 0.0045 | 0.0168 | 102.8336
2017Q2 | 53164725.076 | 0.022167 | 0.61122 | 1.0728 | ... | 97.6873 | 0.007 | 0.017433 | 103.4761
2017Q3 | 53451779.0342 | 0.022367 | 0.61014 | 1.0758 | ... | 97.8137 | 0.0095 | 0.013133 | 103.5137
2017Q4 | 53601437.7291 | 0.020867 | 0.60834 | 1.081 | ... | 97.4819 | 0.011533 | 0.0109 | 104.3091
2018Q1 | 53960814.0875 | 0.019 | 0.60933 | 1.0882 | ... | 97.4234 | 0.012033 | 0.011667 | 104.1112
2018Q2 | 53931032.9449 | 0.0171 | 0.60721 | 1.0937 | ... | 97.5643 | 0.014467 | 0.013133 | 104.5487
2018Q3 | 54343965.1391 | 0.0186 | 0.60591 | 1.1027 | ... | 97.8751 | 0.017367 | 0.011833 | 103.7128
>>
Recursive definitions for new time series are also possible. For
instance one can create a sample from an ARMA(1,1) stochastic process
as follows:
>> e = dseries(randn(100, 1), '2000Q1', 'e', '\varepsilon');
>> y = dseries(zeros(100, 1), '2000Q1', 'y');
>> from 2000Q2 to 2024Q4 do y(t)=.9*y(t-1)+e(t)-.4*e(t-1);
>> y
y is a dseries object:
| y
2000Q1 | 0
2000Q2 | -0.95221
2000Q3 | -0.6294
2000Q4 | -1.8935
2001Q1 | -1.1536
2001Q2 | -1.5905
2001Q3 | 0.97056
2001Q4 | 1.1409
2002Q1 | -1.9255
2002Q2 | -0.29287
|
2022Q2 | -1.4683
2022Q3 | -1.3758
2022Q4 | -1.2218
2023Q1 | -0.98145
2023Q2 | -0.96542
2023Q3 | -0.23203
2023Q4 | -0.34404
2024Q1 | 1.4606
2024Q2 | 0.901
2024Q3 | 2.4906
2024Q4 | 0.79661
>>
Any univariate nonlinear recursive model can be simulated with this approach.