## Info

Figure 7-5: Lotus 1-2-3® template for improving cost estimation.

Figure 7-5: Lotus 1-2-3® template for improving cost estimation.

the tracking signal (TS). The printout* of such a Lotus 1-2-3® spreadsheet is shown"» in Figure 7-5. Appendix E and references 111 and 26) include information on proba-bility, statistics, and forecasting.

Figure 7-5 assumes that for each period (Column A) someone has made an esti-,^ mate of a variable (Column B), and that the actual value of that variable is, sooner^ or later, known (Column C). (It should be noted that Column A need not be time pe-jj riods. This column simply counts the number of estimates made and links esti^ mates with their respective actuals.) Column D calculates the difference between the actual value, A(t), and the estimate or forecast for that period, F(t). Column 1 contains the absolute value of that difference. We can now calculate a statistic^ known as the mean absolute deviation (MAD).

As the information in Row 3 of the spreadsheet shows:

MAD = I(IA(t)-F(t)l)/n where n is the number of differences. The MAD is therefore the arithmetic average of the absolute values of the differences—the mean absolute deviation.

Students of statistics may note that the MAD has certain logical similarities to the standard deviation. Assuming that the forecast errors are normally distributed, the MAD is approximately 80 percent of a standard deviation (see 1111 and else-

•The data for Figures 7-5 and 7-6 were prepared using a Lotus 1-2-3® spreadsheet, transferred to Excel®, and printed. Any of the common spreadsheet programs can easily handle all of the calculations shown in this chapter. Almost all will accept the formulas and calculations from any of the others.