In: Math
Run a multiple regression with trend and seasonal; forecast the next 12 months.
year | Month | Crates |
1999 | Jan | 20400 |
Feb | 13600 | |
Mar | 17000 | |
Apr | 30600 | |
May | 23800 | |
Jun | 17000 | |
Jul | 27200 | |
Aug | 30600 | |
Sep | 34000 | |
Oct | 45900 | |
Nov | 40800 | |
Dec | 30600 | |
2000 | Jan | 13600 |
Feb | 23800 | |
Mar | 30600 | |
Apr | 25500 | |
May | 27200 | |
Jun | 30600 | |
Jul | 23800 | |
Aug | 47600 | |
Sep | 37400 | |
Oct | 45900 | |
Nov | 44200 | |
Dec | 17000 | |
2001 | Jan | 20400 |
Feb | 13600 | |
Mar | 30600 | |
Apr | 22100 | |
May | 23800 | |
Jun | 30600 | |
Jul | 28900 | |
Aug | 34000 | |
Sep | 42500 | |
Oct | 47600 | |
Nov | 30600 | |
Dec | 30600 | |
2002 | Jan | 25500 |
Feb | 20400 | |
Mar | 23800 | |
Apr | 30600 | |
May | 25500 | |
Jun | 30600 | |
Jul | 34000 | |
Aug | 37400 | |
Sep | 44200 | |
Oct | 47600 | |
Nov | 34000 | |
Dec | 37400 | |
2003 | Jan | 25500 |
Feb | 37400 | |
Mar | 30600 | |
Apr | 30600 | |
May | 27200 | |
Jun | 34000 | |
Jul | 47600 | |
Aug | 47600 | |
Sep | 34000 | |
Oct | 51000 | |
Nov | 37400 | |
Dec | 47600 |
ANSWER:
Let me first give you the description of each of the dummy variable.
Dummy_Jan= Binary if month is Jan
Dummy_Feb= Binary if month is Feb
Dummy_Mar= Binary if month is Mar
Dummy_Apr= Binary if month is Apr
Dummy_May= Binary if month is May
Dummy_Jun= Binary if month is Jun
Dummy_Jul= Binary if month is Jul
Dummy_Aug= Binary if month is Aug
Dummy_Sep= Binary if month is Sep
Dummy_Oct= Binary if month is Oct
Dummy_Nov= Binary if month is Nov
Dummy_Dec= Binary if month is Dec
Now, after putting the variables in the regression, we get the model as above:
Crates = 32460-11560*Dummy_Jan-10880*Dummy_Feb-6120.00000000001*Dummy_Mar-4759.99999999999*Dummy_Apr-7139.99999999999*Dummy_May-4080*Dummy_Jun-339.999999999997*Dummy_Jul+6800.00000000001*Dummy_Aug+5780*Dummy_Sep+14960*Dummy_Oct+4760.00000000001*Dummy_Nov
Value of F-Statistic = 6.9 an significance = 0.0000008 << 0.05, so at 5% level of significance the model is important.
The R-adj = 52.38% and R-sq = 61.26%, which implies around 62% of the total variability of the dependent variable is explained by the model.