246. M-M-M

Fully multiplicative

ETS(𝑀,𝑀,𝑀)𝑥𝑡=𝑙𝑡1𝑏𝑡1𝑠𝑡𝑚(1+𝜀𝑡)𝑙𝑡=𝑙𝑡1𝑏𝑡1(1+𝛼𝜀𝑡)𝑏𝑡=𝑏𝑡1(1+𝛽𝜀𝑡)𝑠𝑡=𝑠𝑡𝑚(1+𝛾𝜀𝑡)𝑥̂𝑡+|𝑡=𝑙𝑡𝑏𝑡𝑠𝑡+𝑚+
Example: ETS(𝑀,𝑀,𝑀)

Given

  • Smoothing parameters: 𝛼=0.5, 𝛽=0.4, 𝛾=0.2
  • Initial states: 𝑙0=12, 𝑏0=1, (𝑠3,𝑠2,𝑠1,𝑠0)=(1.2,1,0.8,1), seasonal period 𝑚=4
  • Data:
𝑡12345678910111213141516
x𝑡121081114129131614111518161317

Step 1 — formula

Observation:

𝑥𝑡=(𝑙𝑡1𝑏𝑡1𝑠𝑡𝑚)(1+𝜀𝑡)

Conditional mean (one-step-ahead forecast 𝑥̂𝑡|𝑡1=𝜇𝑡):

𝜇𝑡=𝑙𝑡1𝑏𝑡1𝑠𝑡𝑚

Innovation:

𝜀𝑡=(𝑥𝑡𝜇𝑡)/𝜇𝑡

State updates:

𝑙𝑡=𝑙𝑡1𝑏𝑡1(1+𝛼𝜀𝑡)𝑏𝑡=𝑏𝑡1(1+𝛽𝜀𝑡)𝑠𝑡=𝑠𝑡𝑚(1+𝛾𝜀𝑡)

Forecast steps ahead from time 𝑡 (using current-period states):

𝑥̂𝑡+|𝑡=𝑙𝑡𝑏𝑡𝑠𝑡+𝑚+

where {1,2,3,} is the forecast horizon (how many steps ahead); 𝑚+=((1)mod𝑚)+1 picks the right seasonal slot for the period steps ahead (cycles through 1,2,,𝑚).

Step 2 — apply at 𝑡=1

𝜇1=1211.2=14.4𝜀1=(𝑥1𝜇1)/𝜇1=(1214.4)/14.4=0.1667𝑙1=121(1+0.5(0.1667))=11𝑏1=1(1+0.4(0.1667))=0.9333𝑠1=1.2(1+0.2(0.1667))=1.16

Step 3 — iterate

Each column header is the equation that produced its values. Values rounded to 4 decimal places; arithmetic performed at full precision.

𝑡𝑥𝑡𝜇𝑡=𝑙𝑡1𝑏𝑡1𝑠𝑡𝑚𝜀𝑡𝑙𝑡=𝑙𝑡1𝑏𝑡1(1+𝛼𝜀𝑡)𝑏𝑡=𝑏𝑡1(1+𝛽𝜀𝑡)𝑠𝑡=𝑠𝑡𝑚(1+𝛾𝜀𝑡)
11214.40.1667110.93331.16
21010.26670.02610.13330.92360.9948
387.48760.06849.67980.94890.8109
4119.18530.197610.09271.02391.0395
51411.98740.167911.20151.09271.199
61212.17590.014412.15111.08640.9919
7910.70490.159312.14931.01720.7851
81312.84590.01212.43171.0221.042
91615.23340.050313.02531.04261.211
101413.47060.039313.84711.0590.9997
111111.51290.044614.33731.04010.7781
121515.53890.034714.65391.02571.0348
131818.2020.011114.94691.02111.2083
141615.25870.048615.63361.0411.0094
151312.66340.026616.49061.0520.7823
161717.95230.05316.88881.02971.0238