W02 Forecasting and Estimation

Question 1

Faulty Forecasts

Answer 1

• over-forecast (expecting too high)

- under-forecast (expecting too low)

Question 2

Forecast Data

Answer 2

Transaction Data
Controls Data
Historical Forecast Data
Auxiliary Data

Question 3

Forecasting Methods

Answer 3

Structural vs Time-Series Methods

Question 4

Structural FC Methods

Answer 4

descriptive
level, trend, seasonality
exponential smoothing etc

Question 5

Time Series FC Methods

3 steps
difference to strucutral FC

Answer 5

model underlying dynamic system
that generates data over time
e.g. auto regressive process, ARIMA

1 hypothesis about data-generating process
2 estimate process paramters
3 apply forecasting method for model

can exploit correlations in data

Question 6

Estimation vs Forecasting

Answer 6

descriptive - predictive

past - future

Question 7

Estimation Approaches (non-parametric, parametric)

Answer 7

no particular distribution or model vs underlying distribution

shape of observations vs model paramters

Question 8

Estimation Objectives

Answer 8

unbiased (expected value of estimator equals actual value)

efficient (unbiased and lowest possible variance)

consistent (convergence to true value with increasing sample size)

Question 9

Estimation Challenges

Answer 9

Endogeneity
Constrained Observations
Small Number of observations
Biased Data

Question 10

Regression Estimators are….

Answer 10

…causal predictions

+abrupt changes can easily be explained
+explanatory value

-hard to find all relevant variables and relationships

Question 11

Estimation Maximisation

Answer 11

1 E-Step (estimate mean and sd)
2 test for Convergence
3 M-Step replace constrained obs with ML-estimators