Contents:

• Time Series data
• Linear regression

Time Series data

Time series data can indicate the movement of a variable over years, quarters, months or another time period. The way you describe this kind of data dictates the quality of analysis you are able to carry out. It is not enough to make comments like 'the value increases for the first period before falling back towards the end'. You need to look at the following:

Trend - Over the total period has the value of the item increased, decreased or remained stable?

Key Moments - Does the item change consistently over the period or are there some times when the value is rising and periods when it is falling?

Magnitude - Specific details about the changes over the time period considered, given in nominal or percentage terms.

Seasonality - Is the data seasonally adjusted? Often with economic series of data there is a significant seasonal effect, as spending on the item for example rises or falls according to the time of year. If the data has been adjusted to take care of the seasonal effects, then the longer term trend can be identified more clearly.

Hypothesis - To show your understanding of the theories underlying the data, you need to think carefully about whether the data under consideration sheds light on any particular part of the topic or any particular hypothesis. When you have been given data to analyse as part of a project, you need to look at it in light of your course notes and think if the data illustrates a particular connection. It may be that the data shows that a theory held to be true in textbooks, is in fact rarely evidenced in practice.

There is an illustration available of time series analysis.

[Top]

Linear regression:

The word 'regression' literally means 'a move backwards', but in statistics it helps if we see it as meaning 'a move towards'. What we are moving towards exactly is a way of explaining the behaviour of variables on which we have gathered data. Linear regression assumes that there is a graph line, for which a linear equation is found, that moves towards the observed data to the greatest extent.

It is possible that the observed data you have gathered has values for the x and y coordinates that correspond exactly to those of a single straight line. But it is much likelier that there will be several observations whose coordinates do not conform to a straight line. Regression, in this case, can be thought of as an averaging process during which the best straight line is fitted.

There is an illustration available of linear regression.

[Top]