Time series analysis

A time series is a collection of observations of well-defined data items obtained through repeated measurements over time.

Time series descriptions separate time components in terms of trends (long-term direction), seasonal variation (systematic, calendar-related movements) and irregular cycles (unsystematic, short-term fluctuations). Time series are frequently plotted on line charts or bar charts.

There are two main goals in time series analysis:

  1. identifying the nature of the phenomenon represented by the sequence of observations, and
  2. forecasting (predicting future values of the time series variable).

Both of these goals require that a pattern of observed time series data is identified and more or less formally described. Time series analysis requires that you have at least twenty or so observations.

Examples

Seasonal Calendars

"Purpose: To explore and record data for distinct time periods (per season, year, month or even week) to show cyclical changes over time. From an M&E perspective, calendars can help, for example, to assess if bottlenecks that occurred regularly are being resolved or not, whether these are attributable to the project and when certain performance questions or indicators are best monitored or evaluated.

How to: It is important to clarify with those involved whether calendars will monitor changes between weeks, months, seasons, or years. This will depend on the indicators that have been selected and the rate at which they change.

  1. Construct the calendar either to depict one or several years, or the minimum number of months or seasons over which monitoring is intended to occur. The calendar can be represented either horizontally or as a circle, though the latter can become messy to read if many indicators are being monitored. Circular calendars are not suited for multi-year trend analysis.
  2. The calendar itself can be used to gather the data in some cases. For example, at weekly or monthly staff meetings, when the tasks completed in the past month are discussed, these can be recorded immediately onto the calendar. Alternatively, if data are gathered through other means, then for each time interval for which data is gathered, the correct amount can be filled in, thus using the calendar as a recording format.

A group discussion variant on this process is to divide participants into groups. Each group selects one or two "key informants", who may have relevant expertise, to be interviewed by the rest of the group. Based on this information, each group then makes a diagram to illustrate trends and changes in those activities and/or events over the time interval of interest. These are then presented to the whole group for discussion.

After several data entries, the calendar will show variations over time and so stimulate discussions to understand what the changes are and why they are occurring. By monitoring various types of changes simultaneously in one seasonal calendar or trend chart, certain patterns may become apparent such as how heavy work periods may occur during periods of indebtedness, illness and lower attendance at group meetings. Data can also be differentiated according to age and gender. However, the relevance of such variations will depend entirely on what it is that you want to monitor.

"Daily Routines" Variation

A variation on this option is to depict daily routines (or "how do I spend my 24 hours"), thus looking at daily patterns. This is useful for assessing key bottlenecks in daily tasks and how they can be overcome, or for making quantitative assessments of labour and inputs needed for daily tasks. Comparisons are made between the current situation and previous diagrams to identify how changes that have been introduced affect routines."

Source: (Guijt & Woodhill, 2002)

Advice for choosing this method

  • "The calendar option is ideal for monitoring over specific time periods, such as per season. Seasonal calendars that include a range of indicators can reveal how different patterns of change are linked and can be good for discussing causality of certain changes. Seasonal changes are particularly important for rural areas. They may significantly affect labour, water supplies, disease, food and income."
  • "However, as with historical trends/timelines, seasonal calendars do not necessarily present accurate data. Cross-checking through direct measurement of, for example, time used to fetch water or incidence of diseases may be needed, depending on the accuracy you need."

Source: (Guijt & Woodhill, 2002)

Advice for using this method

  • "If using this option with a group of people, it may be difficult to reach consensus on a "typical" or "average" calendar (particularly when it comes to daily routines). It might be best for each person to do one individually and then analyse the different routines together, or to select one or two individuals in the group as laid out in the second part of Step 3. Care must then be taken to limit biases in the sample."

Source: (Guijt & Woodhill, 2002)

Resource

Guijt, I., & Woodhill, J. International Fund for Agricultural Development (IFAD), Office of Evaluation Studies. (2002). Managing for impact in rural development: A guide for project M & E, annex d.5. Retrieved from website: http://www.ifad.org/evaluation/guide/annexd/d.htm

https://en.wikipedia.org/wiki/Time_series

StatSoft, Inc. (2011). Time series analysis. In: Electronic Statistics Textbook: http://www.statsoft.com/textbook/time-series-analysis/#1general (archived link)

William M.K. Trochim (2006). Time in Research. In: Research Options Knowledge Base: http://www.socialresearchmethods.net/kb/timedim.php

Expand to view all resources related to 'Time series analysis'

'Time series analysis' is referenced in: