Choose a time-series data set that interests you
(ie, a data value is measured over different periods of time).
Some examples may be: closing price of a stock, weekly sales of a company,
a country’s GNP in different years, your height vs age, etc…
The data should be measured over roughly 12-36 periods.
List the periods and data values in a list.
Plot the data points vs periods (with periods on the horizontal axis)
using a Scatter plot with Smooth Lines and Markers (see Excel).
Find the best-fit (regression) trendline with Excel.
Put the trendline onto your graph, along with the best-fit equation
and the coefficient of determination (R^2)
(as was done in class on Tuesday).
This should all be on one page.
On a separate page:
Find the best-fit quadratic and cubic (polynomial) models.
Plot the connected data points with the quadratic model,
the equation and the coefficient of determination (R^2).
Then do so in a separate graph also with the cubic model.
Find the best-fit exponential model;
also find the best-fit logarithmic model.
Plot the connected data points with this model,
the equation and the coefficient of determination (R^2).
Which of these models seems to fit your data better?
On a third page:
Find the three-period simple moving average.
Plot the data values with this moving average.
(Note that this can be done readily in Excel.)
Find a three-period weighted moving average
(you may decide on the weights, but recall
that the weights must all add up to 1).
In a separate graph, plot the data points with this moving average.
Note that you will have to do this by plotting the two series
(both the original series and the moving average) on one graph.
To do so, list the periods in one Excel column, then the data
in a second adjacent column and the moving average in a third;
highlight all columns then insert your connected Scatterplot.)
Which moving average model seems to fit your data better?
Lastly, on a fourth page: Do an exponential smoothing.
Use alpha = 0.1 to obtain a smoothed data set
(ie, place more weight on the less recent data).
Plot the data values with this smoothed series.
Note that you will have to do this by plotting the two series
(both the original series and the smoothed data) on one graph.
(as you did above for the data and weighted moving average).
Then use alpha = 0.9 for another smoothed data set
(ie, place more weight on the more recent data) and plot.
Which model seems to fit your data better?
You should all work independently
and hand in analyses for different data sets.
Since much similar analysis was done in class on Tuesday
you can refer to our lecture notes on Forecasting for more info.
Feel free to email me (or check Excel/online) if you have questions.
