Using gret l for Principles of Econometrics, 4th Edition
Data Structures: Time Series
In order to take advantage of gretl’s many built-in functions for analyzing time-series data, one has to declare the data in the set to be a time-series. Since time-series are ordered in time their position relative to the other observations must be maintained. It is, after all, their temporal relationships that make analysis of this kind of data different from cross-sectional analysis.
If the data you have do not already have a proper date to identify the time period in which the observation was collected, then adding one is a good idea. This makes identification of historical periods easier and enhances the information content of graphs considerably. Most of the data sets distributed with your book have been declared to be time-series and contain the relevant dates in the set of variables. However, it is a good idea to know how to add this information yourself and we show how to do so here. Basically you need to identify to gretl that the data are time-series, you need to specify their frequency of observation, and then identify the starting date. As long as there are no ‘holes’ in the data, this should get you the relevant set of dates matched to the periods they are observed.
Before getting to the specific examples from the text, something should be said about how gretl handles dates and times.
Gretl is able to recognize dates as such in imported data if the date strings conform to the following rules. For annual data, you must use 4-digit years. For quarterly data: a 4-digit year, followed by a separator (either a period, a colon, or the letter Q), followed by a 1-digit quarter. Examples: 1997.1, 2002:3, 1947Q1. For monthly data: a 4-digit year, followed by a period or a colon, followed by a two-digit month. Examples: 1997.01, 2002:10.
Gretl allows you to declare time-series annually, monthly, weekly, daily (5, 6, or 7 per week), hourly, decennially, and has a special command for other irregular dates. Its date handling features are reasonably good, but it is not nearly as sophisticated as those found in other software like Stata. On the other hand, for what it does it is much easier to use. It works beautifully with most datasets.
There are two methods of getting your dataset structured as a time-series. The first uses the GUI. Click Data>Dataset structure from the pull-down menu to initiate the data structure wizard. The wizard serves up a series of dialog boxes that help you to define when the observations occur. The first dialog defines the structure: the choices are cross-sectional, time-series, and panel. Choosing time-series brings up a dialog to set the frequency. Choices include: annual, quarterly, monthly, weekly, daily (5, 6, or 7 per week), hourly, decennial, a special command for other irregular dates. Choosing one of these brings up the next dialog that sets the start point. For instance, quarterly data might start at 3rd quarter of 1972. You would enter, 1972:3 in the box. Then the confirmation dialog opens. It reveals how gretl interpreted your choices. You check to see whether the data start and stop when expected. If so, then your data structure is almost certainly correct. If the end date is something other than you expect, then go back and try again. You may have some gaps in the data series that need to be filled in order for the dates and the number of observations to match up. Sometimes things need manual editing due to holidays and such. Be patient and get
Figure 9.2: Check the confirmation box to be sure the expected time periods are given.
this right, otherwise you may end up having to redo you analysis. Figure 9.1 shows the first three dialog boxes for defining time-series structure. The last box (Figure 9.2) confirms that the series starts in 1960:1 and ends in 2009:4.
The setobs command is used to accomplish the same thing from the console or in a script. The syntax is summarized
Basically you define the periodicity and when the series starts. Then the options are used to indicate what the actual structure is (e. g., time-series). Some examples are found in Table 9.1.