# Thread: Help with Time series

1. ## Help with Time series

Hi Guys,

Could anyone help me with time series in stats?
I have a homework to do but cant determine whether im supposed to use an additive model or a multiplicative one..

Attached is a copy of the graph of data ive got to work with..
Thank you very muchx

Any help or suggestion will be welcomed and appreciated guys.

Cheers,
Amazed

2. Originally Posted by Amazed
Hi Guys,

Could anyone help me with time series in stats?
I have a homework to do but cant determine whether im supposed to use an additive model or a multiplicative one..

Attached is a copy of the graph of data ive got to work with..
Thank you very muchx

Any help or suggestion will be welcomed and appreciated guys.

Cheers,
Amazed
It looks to me as though a linear trend with an additive seasonality signal will be a good model.

RonL

3. Hiya,

Thanks very much Captain...but is there not an increase seasonality trend in the graph? Im am bit confused..

Thanksxxx

4. Originally Posted by Amazed
Hiya,

Thanks very much Captain...but is there not an increase seasonality trend in the graph? Im am bit confused..

Thanksxxx
Not as far as I can see, but without the data I can't subtract off the trend to check, but the seasonality looks stationary to me.

RonL

5. hiya,

If the datas available are in months, should i seperate them into quarters?(refer to the graph attached in original msg) I have to do them by hand and it seems to be a long procedure...
i only have done example with quarterly months and dunno how to do with monthly data...

I really appreciate ya help Captainblack and all others who viewed the thread.Ty

Cheers,
Amazed

6. Originally Posted by Amazed
hiya,

If the datas available are in months, should i seperate them into quarters?(refer to the graph attached in original msg) I have to do them by hand and it seems to be a long procedure...
i only have done example with quarterly months and dunno how to do with monthly data...
I would keep then in months, and use a period of 12 from the seasonality.

RonL

7. Originally Posted by Amazed
Hi Guys,

Could anyone help me with time series in stats?
I have a homework to do but cant determine whether im supposed to use an additive model or a multiplicative one..

Attached is a copy of the graph of data ive got to work with..
Thank you very muchx

Any help or suggestion will be welcomed and appreciated guys.

Cheers,
Amazed
Attached is a spreadsheet showing the detrending of the underlying data, the extraction of the seasonality,
and the reconstitution of the trend and seasonality, and the corresponding residuals. This is all assuming an

sales = a*period + c + f(period)

where f is a function of period which is perioodic with period 12 periods.

The main plots are shown below:

Note the residual plot shows a linear trend but with a low-ish value of R^2.

RonL

8. Originally Posted by CaptainBlack
Attached is a spreadsheet showing the detrending of the underlying data, the extraction of the seasonality,
and the reconstitution of the trend and seasonality, and the corresponding residuals. This is all assuming an

sales = a*period + c + f(period)

where f is a function of period which is perioodic with period 12 periods.

The main plots are shown below:

Note the residual plot shows a linear trend but with a low-ish value of R^2.

RonL
Using the model in the previos post to perdict the next 12 months figures
gives the prediction shown in the attachment:

The first col (after the row number) is the period, the next is the trend, the
third the seasonality, and the final col the trend plus the seasonality which is
the prediction.

From the residual plot we can see that we can expect the error in the predictions
to be less than ~20, and usualy less than ~10.

RonL