turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Forecasting
- /
- Forcasting volume by categorical variables

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

11-22-2013 04:19 PM

Hello,

I wish to predict volume in every day , given a day,month,year, day in week (sunday/monday/...), and holiday indicator. ( total of 5 categorical variables) - base on history data of course.

basically I want to know the weight each of the variables and their values contribute to the volume.

I want to make that forcast as the volume is not necessarily spread uniqly (linear) throw the month. for example: if I'm at the begining of the month I might see a bad picture of what it may look at the end of the month.

So I'm looking for some beta-hats, or weights for the day/month/day_in_week/holiday.

will be glad to know if there is a proc that will do that.

Thanks in advance

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

11-22-2013 05:04 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

11-22-2013 06:32 PM

Thanks for your answer. But the arima proc will give me the actual volume (by the forcast stament) and I only want to know the weighs of the day+month+year+ day in week+ holiday, so I can put into them my wish/ wanted volume for the future.

So the weights has to be precentage or somthing like that.

For example if I will know that Sunday has influence of 10% and my wish volume for the week is 400 and my actual volume on that Sunday was only 35 I know my volume for that day is not good. I hope I cleard myself better this time.

Thanks,

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

11-25-2013 03:38 PM

Hello -

Both ARIMA and UCM will not only give you forecasted values, but much more.

A nice feature of UCM is that it can "decompose" forecasts into their components - which will allow you to do the "what-if" type of scenario you are talking about.

Below you will find an example - hope this helps.

Thanks,

Udo

*first create future values for the masonry variable of sashelp.workers;

proc esm data=sashelp.workers out=masonry plot=forecasts;

id date interval=month;

forecast masonry / method=winters;

run;

*merge the 2 new data sets - note that for masonry it features forecasted values, which you can change later to assess the impact;

data total;

merge sashelp.workers(drop=masonry) masonry;

by date;

run;

*run UCM with mansonry, irregular and level (which is also called trend in UCM) as compoents;

*check out the outfor data set, which features the components of the forecasts;

*outest features the parameter estimates;

proc ucm data=total plots=smooth(decomp);

id date interval=month;

model electric=masonry;

irregular;

level;

estimate outest=outest;

forecast lead=12 outfor=outfor;

run;