|
Frequently Asked Questions
QUESTION:
How do I get a trial copy of AUTOBOX 5.0 for Windows ?
ANSWER:
You can download a data limited trial version right now from the web site.
This version is fully functional but is limited to data sets that are provided. Please click here. A few people have experienced some slight
start up issues relating to versions of Windows, type of hardware, local language
(German, French, etc. ) . If you have any problems please call and we probably can
clear it up on the phone or arrange to email you a new copy.
QUESTION:
What does the six series limit in a Transfer Function mean?
ANSWER:
A transfer function is simply a more powerful version of an ordinary multiple regression.
When we talk about 6 series in a model , we mean at most 5 inputs. If you start
with a univariate model then AUTOBOX will at most be able to include 5 dummy or
intervention series in its augmentation process. Essentially the identification
of intervention variables leads to a multiple series model. The inclusion of causal
variables and interventions can quickly deplete the 5 open positions.
QUESTION:
When you describe the maximum number of data series in each database, what does
this refer to?
ANSWER:
Essentially it was a way to allow us to allow for us to price the product for
bigger prices for bigger problems.
QUESTION:
Is there a maximum number of databases?
ANSWER:
No. AUTOBOX does not place an upper limit on the number of databases.
QUESTION:
If I buy a basic version (e.g. $395) of your package and I want to upgrade to a
more expensive version (e.g. $1195) will I only pay for the difference in price
or will I have to pay for the full amount of the new version?
ANSWER:
AFS does not encourage customers to purchase inexpensive versions and move up. Our
policy is that we will provide a full credit toward any trade up or trade-in for
the first 30 days. For tradeins between 30 and 60 days we will provide a 50% credit.
After that you would have to pay the full price. Thus if you bought the $395 version
and decided within 30 to 60 days that you had made a mistake and wanted a $1195
version, you would have to pay an additional $1000 ( $1195 less the 50% tradein).
It's best to fully understand your needs. We will be glad to help you and suggest
the program that best meets your needs. Call us at 215-675-0652 and we will
explain this to you in full.
QUESTION:
What does the user need to supply to build models with Autobox?
ANSWER:
The data, the time period, the periodicity of the data(e.g. monthly), and the type
of data (stochastic(random) or deterministic(dummy variable).
QUESTION:
How many causal variables can be used?
ANSWER:
The user can specify up to 29 independent variables. Autobox will identify intervention
variables which are included as part of the 29 maximum ( smaller versions are limited
to 5 inputs ).
QUESTION:
How many observations can be used in an Autobox version 5.0 model?
ANSWER:
There can be a maximum of 2000 observations for any given analysis. It would be
most unlikely that a model's parameters would remain constant over that time let
alone the model. Autobox's "Test For Constancy of Parameters" will identify
if the parameters have changed and at the user's option to ignore the data prior
to the point in time when the change occurred. Smaller versions are available with
a maximum length of 100 and 300 respectively.
QUESTION:
Are there exception reports that identify high/low series?
ANSWER:
Yes, it can detect level shifts, pulses, seasonal pulses, changes in variances and
local time trends. These tools provide the user with an early warning system that
highlights which series are experiencing significant changes.
QUESTION:
Can Autobox store model forms for future use?
ANSWER:
Yes, the system can store model forms. These models can be used as either starting
points for the next period's processing or can be used for the forecasting phase
thereby reducing job time.
QUESTION:
Can Autobox include promotional variables in demand forecasting applications?
ANSWER:
Yes, the user can specify dummy variables into the modeling process. Autobox can
identify lead and lag relationships due to the promotion. If the user fails to include
the promotion the system would most likely identify the omission and account for
it through intervention detection. Promotional variables can also be entered as
quantitative as compared to qualitative variables, depending on the user's needs.
QUESTION:
How is missing data and abnormal data handled?
ANSWER:
Intervention detection will correct for any unusual data. AUTOBOX will replace missing
observations omitted from the initial importation by developing smoothed values
via interpolation.
QUESTION:
Does Autobox provide forecast accuracy measurement statistics?
ANSWER:
Yes, MAPE,MAD,BIAS etc.
QUESTION:
Can Autobox use exponential smoothing/Holt-Winters/etc?
ANSWER:
Yes, but these are particular forms of ARIMA which have model forms that are "assumed"
whereas TF and ARIMA modeling procedures make no assumptions about the model form
and generate the model parameters solely on the pattern inherent in the data.
QUESTION:
Can the "forecast engine" that drives Autobox be integrated into existing
systems?
ANSWER:
Yes, this portability can allow users to integrate Autobox into production planning
and decision support systems. We have a DLL that can be called using C
code that we have a available.
QUESTION:
Can data be imported/exported?
ANSWER:
Yes, Access, ASCII and dBase formats are recognized.
QUESTION:
Can the system perform in a batch mode?
ANSWER:
Yes, the user can select many series to execute. The user can also choose to run
interactive (one series at a time).
QUESTION:
What platforms can the system run?
ANSWER:
Autobox can run on all platforms except MAC's. The program is Fortran based which
is portable.
QUESTION:
Can it simulate what-if scenarios?
ANSWER:
Yes, the user can vary the future values of the independent variables and re-model
to see the effect of different scenarios.
QUESTION:
Are there graphs to view the analysis?
ANSWER:
Yes, graphing is sometimes the best way for humans to visualize relationships with
the actuals and forecasted data. Statistics are a numerical way of describing data
and relationships between datasets. Often a combination of the two descriptive methods
(graphical and statistical) can help an analyst build a superior model.
QUESTION:
Is there on-line help/tutorials?
ANSWER:
Yes, we have tried to provide help for terms, usage and every selection that you
will have. We have also provided our a tutorial to provide more "hands-on"
training.
QUESTION:
Does AFS provide training for Autobox?
ANSWER:
Yes, call for details.
QUESTION:
Is there a help desk?
ANSWER:
Yes, it's open from 8am-8pm (EST) Monday though Friday, though you can probably
catch us in on some weekends ;). Dial 215-675-0652.
QUESTION:
Does the company utilize a remote sales force?
ANSWER:
Yes, there are 8 representatives abroad. We are looking for new strategic alliances.
Contact us if you are interested.
QUESTION:
What is the future of Autobox?
ANSWER:
The next utilization of a forecasting model is to use it for its optimization value.
Current Manufacturing goals are to minimize costs given volume projection levels.
These projections are functions of Price and Promotion. The need is to be able to
solve for the optimal combinations of inputs in order to maximize profit rather
than to minimize manufacturing costs or maximize sales. Autobox expects to be able
to one day offer this optimization (backsolving) capability.
QUESTION:
What measures does AUTOBOX provide for actually measuring forecast performance?
ANSWER:
AUTOBOX provides a comprehensive summary of forecasting performance and allows for
forecast errors to be tracked by origin and by lead time. In a sentence one 12 period
out forecast error is not the same as twelve 1 period out forecast errors. Kind
of obvious but most people never think of tracking performance from different origins.
|
ACTUAL DATA
& FOUR PERIOD OUT FORECASTS FROM SIX ORIGINS (LEAD TIME = 4 ; ORIGINS = 6)
|
ACTUAL
|
397
|
378
|
472
|
370
|
395
|
427
|
ORIGIN\DATE
|
1984/3
|
1984/4
|
1984/5
|
1984/6
|
1984/7
|
1984/8
|
1984/2
|
308
|
328
|
399
|
355
|
1984/3
|
|
347
|
472
|
387
|
431
|
1984/4
|
|
|
396
|
404
|
426
|
439
|
1984/5
|
|
|
|
421
|
436
|
441
|
1984/6
|
|
|
|
|
444
|
448
|
1984/7
|
|
|
|
|
|
444
|
|
Note: We have 6 estimates of a one-period out forecast error
Note: We have 5 estimates of a two-period out forecast error
Note: We have 4 estimates of a three-period out forecast error
Note: We have 3 estimates of a four-period out forecast error
Measures To Assess Forecast Model Performance
The above table contains all the raw data necessary to assess how predictable the
future is for alternative lead times. The point is simple and profound. Forecasting
accuracy from a single launch point generates correlated forecast errors. Forecast
error analyses should use a number of different origins and a number of lead times.
To reiterate, a set of historical values are used to perform some modeling activity,
be it automatic or not, in order to come up with a model and a set of coefficients.
These are then fixed and each of the withheld observations are then used as the
launching point for a new set of forecasts. With this approach the new observations
don't fully participate in the modeling process and only affect the forecast but
not the model form nor its parameters.
Typical Output Tables From AUTOBOX
VALUES ARE IN TERMS OF THE ORIGINAL
METRIC
|
Number of Actuals
|
3
|
Forecast Mean Deviation (Bias)
|
469.327
|
Forecast Mean Percent Error
|
1.20294
|
Forecast Mean Absolute Deviation
|
1601.47
|
Forecast Mean Absolute % Error
|
4.26474
|
Forecast Variance (Precision)
|
.39e+077
|
Forecast Bias Squared (Reliability)
|
220268
|
Forecast Mean Square Error (Accuracy)
|
.41E+07
|
Relative Absolute Error
|
.3969
|
Typical Output Tables From AUTOBOX (MORE)
|
|
Lead Time
|
MEAN DEVIATION (BIAS)
|
MEAN % ERROR
|
MEAN ABSOLUTE DEVIATION
|
MEAN ABSOLUTE % ERROR
|
1
|
.18E+04
|
4.49
|
.41E+04
|
10.75
|
2
|
.22E+04
|
5.49
|
.36E+04
|
12.35
|
3
|
.38E+04
|
5.49
|
.48E+04
|
15.75
|
4
|
.15E+04
|
2.29
|
.21E+04
|
5.75
|
Typical Output Tables From AUTOBOX (MORE)
|
|
Lead Time
|
VARIANCE (PRECISION)
|
BIAS SQUARED (RELIABILITY)
|
MEAN SQUARE ERROR (ACCURACY)
|
RELATIVE ABSOLUTE ERROR
|
1
|
.23E+08
|
.32E+07
|
.26E+08
|
.54
|
2
|
.24E+08
|
.36E+07
|
.24E+08
|
.62
|
3
|
.24E+08
|
.56E+07
|
.34E+08
|
.39
|
4
|
.14E+08
|
.56E+07
|
.44E+08
|
.34
|
Click here to return to the home page
P.O. Box 563
Hatboro, PA 19040
Tel: (215) 675-0652
Fax: (215) 672-2534
sales@autobox.com
|