Week 6 Lecture 3 Forecasting Flashcards
Under what circumstances can linear regression be used?
When the data is linear and constant
As opposed to the weight given to Moving Averages, the weight given in Exponential smoothing…
Changes and is determined by Alpha
Under what circumstances should ‘Single Exponential Smoothing’ be used?
When the data is non-trended and non-seasonal
The alpha used in Exponential Smooting is…
Not calculated. It is an estimate set beforehand.
Between what values should Alpha lie?
Between 0 and 1
What is the formula used in Single Exponential Smoothing for the calculation of St?
Alpha x Yt + (1-Alpha) x St-1
Under single exponential smoothing, the forecast for year 2 is always the (………) of year one?
Sales/Data point
In relation to significance, a lower alpha…
Gives more significance to older data.
In relation to significance, a higher alpha gives…
Less significance to old data
A smaller alpha results in what sort of time-series graph line? And why is this the case?
A smoother line.
This is the case because the graph takes longer to react to changes i the data
What is the key consideration when setting the alpha?
Is the old data significant? If it is we choose a lower alpha, if it is not, we choose a higher alpha.
In single exponential smoothing, a line on a graph representing the quick recognition of changes in data points is likely to be more…
Erratic
Although it is not possible in practise, an alpha of 0 would result in what? And why is this the case?
A straight line along the point of the first data entry.
This is because it gives absolute significance to the old data.
What is the idea range for an alpha to lie?
Between 0.1 and. 0.3
What is the Alpha also known as?
The smoothing constant
Under single exponential smoothing, the weight of all data points ad up to what value?
1
Under what circumstances is Holts method used?
When there is a trend in the data.
Under Holts method, 2 equations are needed, what are they?
St = Alpha(Y1) + (1-Alpha) (St-1+Bt-1)
Bt = Beta(St - St-1) + (1-Beta)bt-1
In the Holt model, what does Beta represent?
The smoothing level of the estimated trend.
Beta is either given or is calculated how?
By taking an estimate from the first 3 data points.
For example: (3rd point - 1st point)/2
When making forecasts using the Holts method, how are they calculated?
The St of the last period + the Bt of the last period.
Between what values can Beta be?
0 and 1
