Hi k9queen!
These are the answers:
a) The procedure to get from the original data to the seasonal indices
is explained in a previous question I answered for you:
"The idea is the following: compute the CMA for each quarter and
compare the actual value for demand each quarter with this average.
Divide each actual value by its corresponding centered moving average.
The point of doing this is that you're comparing how different is the
current quarter from the ones "around" it (through the CMA). If the
actual value is higher than the CMA, it means that there is a positive
seasonal effect in the current quarter. For example, mall sales tend
to be substantially higher in the 4th quarter because of Christmas
shopping. You would see that the actual value for sales in the 4th
quarter is higher than its corresponding CMA, because the CMA takes
the average among months that didn't have the Christmas effect.
So, back to the procedure. After dividing each value by its
corresponding CMA you will have several indices. Finally, you define
the seasonal index of quarter 1 as the average among the indices you
got for all 1st quarters; the seasonal index of quarter 2 as the
average among the indices you got for all 2nd quarters, etc. These are
the seasonal indices. A seasonal index greater than 1 implies that
demand that quarter is higher than "normal", while an index smaller
than 1 implies the opposite."
Forecasting... - Question ID: 274754
http://answers.google.com/answers/threadview?id=274754
Once we have the seasonal indices, in order to compute the
deseasonalized data we have to divide each original value by its
corresponding seasonal index. So you would have to divide the 1st
quarter of every year by the seasonal index of Q1, the 2nd quarter of
every year by the seasonal index of Q2, etc. This "smooths" the data.
Since quarters of extremely high sales will have a high seasonal
index; when dividing the actual value by the index we would get a
smaller value , a value that is closer to the sales in recent
quarters.
The formula used to calculate the coefficients of the trend line can
be found in slide 6 of the following Powerpoint presentation (if you
have trouble viewing it, try right clicking on the link and choosing
"Save target as...")
Forecasting
http://www.gsu.edu/~dscgpz/chap2/chapter2.ppt
b) The procedure of computing the forecats is very simple. First of
all, we need to know the number that correspond to each quarter of the
11th year. Let's assume that the 1st quarter of the 1st year was 0,
the 2nd was 1, the 3rd was 2, and so on. This implies that the 1st
quarter of the 11th year would have the time index 44; the 2nd one,
45; the 3rd one, 46; and the 4th one, 47. Next we use this numbers to
compute the deseasonalized trend forecast:
Q1, year 11: 1,000,000 + 25,725*44 = 2,131,900
Q2, year 11: 1,000,000 + 25,725*44 = 2,157,625
Q3, year 11: 1,000,000 + 25,725*44 = 2,183,350
Q4, year 11: 1,000,000 + 25,725*44 = 2,209,075
Since the trend line was calculated on deseasonalized data, in order
to make the final forecast we have to re-seasonalize the trend
forecasts. This is simply to multiply each forecast by the
corresponding seasonal index. So we get:
Q1, year 11: 2,131,900*1.12 = 2,387,728.00
Q2, year 11: 2,157,625*0.87 = 1,877,133.75
Q3, year 11: 2,183,350*1.29 = 2,816,521.50
Q4, year 11: 2,209,075*0.98 = 2,164,893.50
In this way, we get bakc the differences among the quarters (sales
peak in Q3, etc.)
Google search strategy
seasonal adjustment
://www.google.com.ar/search?q=seasonal+adjustment&ie=UTF-8&oe=UTF-8&hl=es&meta=
trend forecasting definition
://www.google.com.ar/search?hl=es&ie=UTF-8&oe=UTF-8&q=trend+forecasting+definition&meta=
I hope this helps! If you have any doubts regarding my answer, please
don't hesitate to request a clarification. Otherwise I await your
rating and final comments.
Best wishes!
elmarto |
