0
$\begingroup$

really appreciate any help, i've been running HoltWinters on data below, using R. Initially i just used

Call:
HoltWinters(x = testTS)

and obtained...

Smoothing parameters:
 alpha: 1
 beta : 0
 gamma: 0

Coefficients:
           [,1]
a   10319.23997
b     156.31002
s1   1715.93795
s2   1203.04212
s3  -2452.57288
s4  -2857.84413
s5  -4193.45622
s6  -3572.14455
s7  -1655.78747
s8   -235.90663
s9     99.81837
s10  1739.35295
s11  5596.80045
s12  4612.76003

but when i re-did and set Gamma to anything other than zero, i get exact same coefficients!?

E.g. setting gamma = 0.5

Call:
HoltWinters(x = testTS, alpha = 1, beta = 0, gamma = 0.5)

Smoothing parameters: alpha: 1 beta : 0 gamma: 0.5

Coefficients: [,1] a 10319.23997 b 156.31002 s1 1715.93795 s2 1203.04212 s3 -2452.57288 s4 -2857.84413 s5 -4193.45622 s6 -3572.14455 s7 -1655.78747 s8 -235.90663 s9 99.81837 s10 1739.35295 s11 5596.80045 s12 4612.76003

am i missing something obvious here? I thought should have ended up with different coefficients. Also, in my initial model when i get alpha=1, beta=0, gamma=0, could someone please give me their thoughts on these values? My data did follow a reasonable steady trend/seasonality up until April 2020, then a major deviation...i'm just not sure whether this sort of dataset should yield beta=0, gamma=0

          Jan   Feb   Mar   Apr   May   Jun   Jul   Aug   Sep   Oct   Nov   Dec
**2015**  5914  7315  8992  7345  5559 11498 15304 14336 11442 11081  7805  7697
**2016**  6454  7162  9214 10783 11234 12944 16084 15629 13402 12703  8947  8265
**2017**  6889  7701  9477 11811 13061 13771 17079 16781 13674 11762  9373  7883
**2018**  7097  7874  9661 12024 14296 14499 16889 16419 13626 11144  8958  8186
**2019**  7443  7873 10072 13267 15350 15327 17680 17331 14094 13495  9973  8709
**2020**  7855  7912  7923  3223  3724  4954 10755 14932 

data trend

dput of data now added:

structure(c(5914, 7315.47, 8992.8, 7345.05, 5559.49, 11498.03, 15304.04, 14336.13, 11442.14, 11081.72, 7805.8, 7697.25, 6454.44, 7162.16, 9214.08, 10783.19, 11234.03, 12944.8, 16084.52, 15629.45, 13402.33, 12703.03, 8947.25, 8265.45, 6889.07, 7701.42, 9477.21, 11811, 13061.04, 13771.66, 17079.3, 16781.33, 13674.74, 11762.96, 9373.96, 7883.29, 7097.42, 7874.26, 9661.75, 12024.28, 14296.55, 14499.77, 16889.92, 16419.96, 13626.99, 11144.69, 8958.76, 8186.51, 7443.21, 7873.78, 10072.69, 13267.98, 15350.17, 15327.43, 17680.14, 17331.35, 14094.49, 13495.49, 9973.39, 8709.81, 7855.86, 7912.77, 7923.18, 3223.22, 3724.24, 4954.16, 10755.62, 14932), .Tsp = c(2015, 2020.58333333333, 12), class = "ts")

$\endgroup$
5
  • $\begingroup$ Can you please edit your post to include the output of dput(testTS)? $\endgroup$ – Stephan Kolassa Sep 4 '20 at 16:21
  • $\begingroup$ Hi Stephan, excuse my ignorance wasnt sure what you meant by dput in your comment? I did edit my post to include 2nd set of coefficients (for HW model with gamma = 0.5). Thank you very much $\endgroup$ – Jimmy Sep 4 '20 at 20:19
  • $\begingroup$ @Jimmy: If you're using R, dput will output the actual data contained in the testTS object. $\endgroup$ – mlofton Sep 5 '20 at 0:08
  • $\begingroup$ Please type dput(testTS) in your R console. R will give you an output. Copy and paste that exact output into your question. Then we can paste that into our R consoles to work with your exact data. See ?dput. $\endgroup$ – Stephan Kolassa Sep 5 '20 at 6:45
  • $\begingroup$ have done the dput now (correctly i hope) and pasted it in. Thanks for your patience, clearly i'm new to this. $\endgroup$ – Jimmy Sep 5 '20 at 11:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.