Professional Documents
Culture Documents
OPIM-274 HW2-Solutions
OPIM-274 HW2-Solutions
OPIM-274 HW2-Solutions
Fall 2022
Homework Assignment 2-Solutions
1) Problem 2.8
The correlation is 0.955. Note that Year 1 actual values serve as forecasts for Year 2. So, we
compute the correlation of forecasts and actuals in the data. As would be expected the
consumption pattern is similar from year to year.
Note: Do not use software built-in functions to find the correlation. Use the formula
and show enough details to identify that you used the formula.
r=
∑ (x i−x )( y i− y ) = 2136591.667 =0.954775
√ ∑ (xi −x)2 ∑ ( y i− y)2 2237794.341
(See attached workbook Electricity_2.8.xlsx for detailed calculations)
2) Problem 2.10
1500
Total Revenues
1000
500
Quarter Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1
Year 2000 2002 2004 2006 2008 2010 2012 2014
120
100
Growth - Absolute
80
60
40
20
Quarter Q1 Q1 Q1 Q1 Q1 Q1 Q1 Q1
Year 2000 2002 2004 2006 2008 2010 2012 2014
1
a. The mean and median are not useful, given the strong trend in the data.
b. There is a clear slowing of percentage growth over time becoming stable around
20% percent. Note the outlier in 2011Q1 as the result of an acquisition.
3) Problem 2.11
The results for Forecast_1 (previous year) and Forecast_2 (long-run average) are
2
The median measures are generally smaller but the patterns are consistent.
a.
Monthly Average Forecasts appear to work the best given that it results in smaller errors
across most measures.
b.
This comparison is not fair. The monthly averages are calculated with all 25 years of
temperatures already factored into the forecast.
c.
The median measures are generally smaller but the patterns are consistent.
d.
The errors for both forecast methods include outliers and thus measures based on
medians are more reliable.
4) NVIDIA
a. Time Series Plot
3
Increasing (nonlinear) trend between September 2021 and mid to late November 2021,
then generally, decreasing with fluctuations.
f. Forecast 1 (Random Walk forecast) shows consistently better performance based on all
the measures (i.e., error measures are smaller for Forecast 1 relative to the ones for
Forecast 2).