Attribute Details Panel
(Page under construction.) This panel shows all the details of
an attribute in Istar.
Click on a gadget and I will explain it.
Row 1 identifies the attribute, with:
- Label.
Leftmost in row 1 is the label (name) of the
attribute, that will appear as text in the attribute box.
- Meaning.
The right half of row 1 is occupied by the
meaning of the attribute, that will flash into the meaning bar
at the top of the control screen when the mouse passes over
this attribute.
In addition, in the middle of row 1 is a wee OK button which
acts exactly like the main one further down the panel. It is
here for your convenience for when you are merely filling in
label and meaning and no other details.
Leftmost in row 2 is the domain of the
attribute, that is, what type of value it is. It has two
parts: the main window in which the name of the domain is
shown, and a wee button to hit with the mouse to alter the
domain.
In the middle area of row 2 is the current
value of the attribute. It might be just a checkbox as
here, or a longer gadget such as a slider for bayesians or
probabilities, a number for an integer or twin numbers as for a
ratio.
Over the right hand side of row 2 you
sometimes get extra information about the value type:
- For bayesians: a slider showing the a-priori value
- For enumerated and ordinal types: the number of elements in
set
These relate to the value.
- Show Value on Easel.
The leftmost button in row 3 is
a checkbox to say whether the current value of the attribute is
to be shown on the main easel. If ticked then the value of
probability, bayesian and proportion attributes is shown as a
horizontal line along the bottom, and this is either solid if
answered and dotted if not. If the attribute is of a boolean
domain, then a line is shown that occupies the right half to
show 'true' and the second left-quarter if 'false'. Other
domains are not shown at present.
- Infinite.
The second button in row 3 is ticked if the
value is infinite (as when dividing by 0). At present not
functional.
- Unknown.
The third button in row 3 is ticked if the
value is unknown and not ticked if known.
- Must Be Known.
You tick the rightmost button in row 3
if the value must be known; that is unknown is not allowed.
Normally used for question attributes, in which case the
'Unknown' button is omitted from the User Question Panel.
This is on the fourth row of the panel from version 1.1 (and
lower down in previous versions).
- Answered.
This checkbox shows whether the attribute
is answered or not.
You cannot alter it internally (though
you can alter the gadget).
Waiting. Shows that an asynchronous attribute has
been asked but its answer has still to arrive. Not properly
implemented yet (v1.1).
Error Indicator. Shows an error has occurred in
calculating the value of this attribute, e.g. that a 'must be
known' attribute is found to have an unknown value. Not yet
fully implemented.
This radio button, over the left side of the panel and nearly
half way down, shows how this attribute's value is to be
derived. There are four options:
- Constant: The value is constant, as given in the value gadget.
- Inferred: (the default) The value is
inferred from antecedent attributes by the
inference method. If this option is selected
but there are no antecedents then it is treated as user-supplied.
- User-supplied: A question is put
to the user, whose text is the question
text if present, or, failing this, the meaning text or the label.
- Async:
This is like user-
supplied, in that a question is put to the user, but
asynchronously. In this case the question is put but the
search for input continues, and the user can answer at any
time. This facility is not likely to be very useful on its
own; it is designed for use when, instead of asking the human
user, Istar will post a request for information on the
Internet.
Inference Method for Attribute. This is a
type gadget lying alongside the Infer
radio button. Its window shows the current method, and hitting
the wee button to its left brings up a list of inference
methods appropriate to the domain of this
attribute.
In this area you sometimes get extra information about the
inference mechanism. For bayesians there are two cutoffs,
lower and upper, whose purpose is to ignore minor antecedent
factors. Suppose a bayesian attribute has a lower cutoff set
at 10% and four antecedents A,B,C,D. A, B are answered with
low values and the weights on C,D are such that however high
they are answered the result cannot exceed 10% then the
attribute is considered answered and C and D are ignored.
Conversely for upper cutoff.
Sometimes called 'Reset Value', 'RV'. This controls the
default value that is shown to the user in a question panel,
that is the value that will be taken if they click 'OK' without
setting the value. (Used to be called 'reset value', as the
value the attribute was reset to before the inference session
began, but in fact it is now only the value shown in the panel
that is set to this; so better called 'question value'.)
Normally the question value is the current value of the
attribute. But sometimes you want to present a certain value
every time, as a kind of default answer. When this is so, set
the left-most checkbox and set the right-most gadget to the
value you wish the user to see in a question panel.
Explanation/Help Text. The long string
gadget preceded by 'X' or 'H' holds explanation or help text.
This is text that is dsiplayed to the user if they hit the
'Help' button.
Can Override. Toggle the checkbox, to
left end of override row, to say whether or not the attribute
can have an override value.
Overridden Status. This checkbox says
whether the attribute is currently overridden. In the present
version of Istar, you can alter it.
Overriding an attribute allows the user to temporarily insert a
value of their choice that takes priority over the value
obtained by the normal derivation. This
might be used, for instance, if the user is to be given the
chance to say that they do not like the value that was
inferred. Normally you would place overridable attributes into
one or more override lists.
Next row: Other information.
- Topics
. Bring up the Topics
panel to allow you to link the attribute to topics of your
choice.
- References
: Not yet used. Will allow you to link
attribute to external references.
- Item
to which this attribute belongs.
- 'See' takes you to the Item
Details Panel.
- DSAP
: Unique internal identifier for the Attribute.
('Data Structure Area Pointer').
This row deals with the visual appearance:
- DSAP of EaselPiece that holds the visual appearance.
- Pens: first is label colour, second is box outline.
- Pattern: line pattern for drawing box: 1 or 0 represents a
pixel in outline, repeating every 16.
- Size: width and height in 1/256 of an inch (approx 0.1 mm)
on 14 inch monitor and standard zooming.
These buttons exit this panel. They include:
Bottom of panel, left for antecedents, right for
consequents, are two lists
with buttons. These are underneath the exit buttons because
anything you do with these lists will not be cancelled if you
hit the 'Cancel' button.
- 'See': Click this to raise the attribute details panel for
the selected attribute.
- 'R': Click this to raise relationship instance panel for
the selected link.
- 'X': Click this to gain explanation of what happens in the
link.
- '1': Click this to make that attribute first in list.
- 'Wts': Click this to bring up the Weights Panel for either the antecedent
or the consequent links. So you can set the links all
together.