Scribd
Upload
105 views3 pages

Jiahui He Bnad Lab8

This report analyzes quarterly sales data using exponential smoothing with λ=0.2 and λ=0.9, seasonal dummy variables, and a time trend variable. Exponential smoothing with λ=0.9 produced lower MAD and MSE, indicating it better fits the data. Seasonal dummies showed sales in Q1 are on average 18 units lower than Q3, while Q2 and Q4 are not significantly different from Q3. The time trend variable suggests sales increase by 0.03 units per day on average. In conclusion, exponential smoothing with λ=0.9 fits the data best, and the seasonal and time variables help explain sales patterns between quarters.

Uploaded by

api-530849574
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
105 views3 pages

Jiahui He Bnad Lab8

This report analyzes quarterly sales data using exponential smoothing with λ=0.2 and λ=0.9, seasonal dummy variables, and a time trend variable. Exponential smoothing with λ=0.9 produced lower MAD and MSE, indicating it better fits the data. Seasonal dummies showed sales in Q1 are on average 18 units lower than Q3, while Q2 and Q4 are not significantly different from Q3. The time trend variable suggests sales increase by 0.03 units per day on average. In conclusion, exponential smoothing with λ=0.9 fits the data best, and the seasonal and time variables help explain sales patterns between quarters.

Uploaded by

api-530849574
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 3

ANALYTICS REPORT

TO: BOSS

FROM: JIAHUI HE

DATE: MAY.5 2020

Introduction
This report develops an exponentially smoothed series using both λ = 0.2 and λ = 0.9 and
uses seasonal dummy variables and a linear “time” variable to compare quarterly sales
against quarter 3 (this quarter was chosen because it has historically been the quarter with
the highest sales). Finally, we conclude that the exponentially smoothed series using λ =
0.9 is better, and figured out the relationship between quarter 1,2, 4 and quarter 3.

Data Analysis

Exponential smoothing:

1. λ =0.2

Sales A(t) Forecast Error |Error| Error^2


70 70 lambda = 0.2
59 67.8 70 -11 11 121 MAD MSE
93 72.84 67.8 25.2 25.2 635.04 45.20012 4122.015
96 77.472 72.84 23.16 23.16 536.3856
145 90.9776 77.472 67.528 67.528 4560.031 Jiahui He
179 108.58208 90.9776 88.0224 88.0224 7747.943
321 151.065664 108.5821 212.4179 212.4179 45121.37

2. λ =0.9

Sales A(t) Forecast Error |Error| Error^2


70 70 lambda = 0.9
59 60.1 70 -11 11 121 MAD MSE
93 89.71 60.1 32.9 32.9 1082.41 41.3119 3938
96 95.371 89.71 6.29 6.29 39.5641
145 140.0371 95.371 49.629 49.629 2463.038 Jiahui He
179 175.10371 140.0371 38.9629 38.9629 1518.108
321 306.410371 175.1037 145.8963 145.8963 21285.73
When λ =0.2, MAD equals to 45.2001, and MSE equals to 4112.015.

When λ =0.9, MAD equals to 41.3119, and MSE equals to 3938.

As 41.3119 < 45.2001 and 3938 < 4122, the second model (λ =0.9) is better.

In general, if you just care about your average error, you will choose the model with the
lowest MAD. And, if you want to avoid large outliers or errors, for example, if one big
mistake could bankrupt your business, then you would want to use the model with the
lowest MSE.

December 22 = 0.9*113+0.1*133.5325 = 115.0533

Seasonal Dummy Variables:

SUMMARY OUTPUT
Jiahui He
Regression Statistics
Multiple R 0.14675321
R Square 0.021536505
Adjusted R Square 0.016860454
Standard Error 71.97326033
Observations 842

ANOVA
df SS MS F Significance F
Regression 4 95432.96158 23858.24 4.605704 0.00110533
Residual 837 4335785.719 5180.15
Total 841 4431218.681

Coefficients Standard Error t Stat P-value Lower 95% Upper 95%


Intercept 75.04668137 6.840559866 10.97084 2.94E-26 61.62001492 88.47334782
Q1 -18.3964355 7.366767209 -2.49722 0.012708 -32.85594295 -3.93692811
Q2 -2.91620957 7.260003887 -0.40168 0.688021 -17.1661617 11.33374256
Q4 -5.58686068 6.618121955 -0.84418 0.398812 -18.57692549 7.403204135
t 0.030126282 0.010340819 2.913336 0.003671 0.009829298 0.050423266

^
Sale=¿ 75.047-18.396Q1-2.916Q2-5.587Q4+0.03t

2
R Square: We are 2.15% of the way to perfectly predicting the E/S using this model.
Standard Error: Predictions of E/S using this model are off by an average of 71.97.

Sales on a day in quarter 1 are 18.40 units less than sales on a day in quarter 3, on
average and all else constant.

As far as we can tell, sales on a day in quarter 2 or quarter 4 are the same as on a day in
quarter 3.

Each day, sales increase by 0.03 units sold, on average and all else constant.

It takes about 33 days (or one month) for sales to increase by an average of 1 sale per
day.

December 23 = 75.047-5.587*1+0.03*843=94.75

Conclusion

To conclude, the exponentially smoothed series with λ = 0.9 is better and the seasonal
dummy variables and a linear “time” variable help figuring the relationship between the
sale of Quarter 3 and other quarters.

You might also like