1. Select a data series that you wish to forecast. Make sure that it has some importance to you relative to business, future occupation or other special interest. Obtain monthly or quarterly time series data from one of the data sources listed in either of these websites: www.rfe.org, www.economagic.com or other approved sources shown in “Data Sources” under Chapter 1 of the Course Home. Note, do not select seasonally adjusted data.
2. You need at least 80 observations. You may have more if you wish. You can refer to the project outline for additional details. Download and save the data as an EXCEL file. EXCEL files can be easily imported or copied into Minitab. The first column should be the date and the second column should be you data. Always label the columns appropriately and include (Y) as part of your data label. Create a citation for the data series including the exact webpage or document page that contained the data.
3. Go back to the same data source or other approved data source. This time you will need to obtain time series data for a minimum of three related “explanatory” variables. We will call these related variables the independent (X) variables. These variables will be monthly or quarterly with dates consistent with your (Y) variable downloaded previously. If your (Y) variable is monthly, your (X) variables must also be monthly. If your (Y) variable is quarterly, your (X) variables will also need to be quarterly. All variables should be from the same time period. You can refer to the project outline for further details. Note, do not select seasonally adjusted data. You want the data for each variable to be unadjusted for seasonality.
(See the note below relative that data that should not be used in the project.)
4. Download and save all your (X) variables as an excel file. Label each column of variable data accordingly. Recall that the first column is the dates and the second column is the selected (Y) variable. The third column will be (X1), the forth column will be (X2) and the fifth column will be (X3) and so forth for additional variables. Be sure to include a name of data in each column label in addition to the appropriate (X) notation. Please ensure that the data observation dates are in sync with the appropriate dates in the first column (C1). Make sure that each Y and X data series has a separate citation. This can be placed after the EXCEL data columns.
5. Copy and paste your dates (C1), (Y) variables (C2) and X variables (C3 to C5, etc.) into the MINTAB worksheet section. Copy your labels as well so that I can understand which variables you are analyzing. Save the EXCEL worksheet with your project name followed by data. It must be included as the last page of the proposal.
6. Run basic (descriptive) statistics from Minitab on each variable beginning with Y. Comment on the mean value, data range and the standard deviation. Is the variation about the mean relatively large or close to the mean value? Remember that large standard deviations about the mean may indicate forecasting difficulty.
7. Run individual “Time Series Plots” on each Y and X variables. Comment on what you see in the time series plot (positive or negative trend, periodic business cycles (lasting more than one year), seasonality (cycle lasting exactly one year). Comment on the X variables similarity or difference from the observations of the Y data.
8. Run “Scatter Plots” of (Y) against each (X) variable separately. Use Minitab and enter the Y variable first then the X variables in turn. Comment on the linear (or lack of linear) nature of each YX variable relationships. The relationships may be strongly (positive or negatively linear), moderately related, weakly related or not linearly related. With three X variables you should have three scatter plots to discuss.
9. Get the “Correlation Matrix” (you can find it under Stat/Basic Statistics/Correlation in Minitab) of (Y) and each of the explanatory (X) variables. You need to tell me which (X) variables are linearly related to (Y) and which (X) variables are not. Also note the correlations between the X variables. They should not have stronger correlations than each X variable correlation to Y. If the XX correlations are stronger you may wish to replace the X variable with the weakest correlation to Y with another. Complete any variable replacements before you submit the proposal.
10. The proposal should include (A) your statement of the forecasting problem. Give a good reason for choosing the Y variable ---“It is important to me because…”. Follow that with (B) statements regarding why you chose each of the independent variables (X) for your analysis. This will be your hypotheses relative to the explanation of why Y changes. Follow that with the analysis of the data in the points 6 and 7 above. Then provide further support of your hypothesis with scatter plots and correlation coefficient findings of each X variable’s relationship to the (Y) variable in points 8 and 9 above.
11. Remember that the proposal should be a formal word document and include the relevant MINITAB output copied and pasted in word as well as the description and source of each variable. Minitab worksheet elements and Minitab graphs can be copied and pasted into the proposal as required. Again make sure to include all of the data and website or document that your data came from as your source.
12. Use 12 point New Times Roman type style double spaced for the proposal. You do not need a cover page for your proposal. Include a title in upper case centered type, the class and section number, the date and your name in upper/lower case centered at the top of the first page. Number your pages and be sure that you have run spell check on the entire document. Any subheadings should be centered upper/lower case type. Most proposals are no longer than 5 pages including 2 pages for the data. Do not include any forecast in the proposal. The proposal essentially sets up the forecast problem that you will address in this course.
This document should be uploaded to me at our designated eCollege Dropbox for the project proposals by midnight 2/08/2013
What to avoid is selecting Y and X variables.
1. Do not use seasonally adjusted data.
2. Make sure there are no missing data values.
3. Use either monthly or quarterly time series data.
4. You must have at least 80 observations of each variable.
5. Do not use Macro variables such as Money Supply, GDP, GNP, CPI, PPI, U.S. Interest Rates, U.S. Unemployment or Employment, U.S. Population for a Y variable. They may be used as X variables, however.
6. Do not select known components of the Y variable as X variables. (E.g. using male unemployment and female unemployment as X variables to forecast total unemployment.)
7. Try to use X variables that have logical connection to Y. The X variable should be useful in explaining variation in the forecast Y variable.
8. Try to use X variables that have a reasonable linear relationship with Y. That is the scatter plots indicate a linear relationship and the XY correlations are .30 or better.
9. If the XX correlations are greater than XY correlations drop the lowest correlated X variable and replace it with another variable that 1) has a significant correlation with Y and has a lower correlation with the other X variables.