Description Objects from the Class Slots Methods Author(s) Examples
Three almost independent points are described in /win/ to define a new node with a non parametric regression style from a data base. (I) the variable names and their natures. (II) The windows to select a subset of prediction individuals, usually based on some covariables (the parents in the Bayesian network perspective); the associated slots start with a s
. (III) The way from which a value is computed to represent the selected set, sometimes called the prediction; the associated slots start with an r
. Be aware that most of the sophistication of the algorithm is linked with the fact that there can be several parents and the prediction be multivariate.
Examples of such objects are rbsb.win?
. The usual way to create a new object of this class is by calling the generic function new("win",swg=...)
.
nat
:A named "character"
providing the natures of the variables with their names (without node indication) by its names. At least one variable is expected when it is not the null /win/.
swg
:A named "numeric"
providing by its names the parent variables (in the format node[variable]
) and by its values their weights to compute the distance determining the subset of candidates. When there are no parents, all individuals are set to the same distance.
skk
:A "numeric(1)"
: the power coefficient in the distance formulae (cannot be negative).
sdi
:A "numeric(2)"
: the minimum and
maximum distances between the data base individuals and the individual to be predicted to belong to the candidate subset.
snb
:A "numeric(2)"
: the minimum and
maximum numbers of individuals in the candidate subset. When less are proposed with the distance criteria a NA
is returned; when more, the closest ones are retained (in case of ties, random drawing is performed).
rty
:A "character(2:3)"
to indicate the way the prediction is done from the selected subset.
rty[1]
precises the way the prediction is done:
'*' each column independently,
'0' the real row, the closest to '*' (according to the defined distance
for the representativeness)
and '1' a real row selected from 'rmo
' variable using
rty[2] prediction type.
rty[2]
precises the type of prediction:
'random' a random draw with possibly a weighting
given by the absolute value of rmo
when rty[1]=='1' or '*'
(if not equal weighting is given to all observations),
'mean' the value of the individual closest in mean, for the variable rmo
when rty[1]=='1'or'*',
'median' like 'mean' but using the median,
'systematic', the frequency of the drawing is given by rk2,
if it does not exist it is set to one;
when rty[1] is '1'or'*', rmo
variable is taken to define
the order of drawing if not the order of the data basis is used;
'quantile', as median but for any probability level indicated in rk2.
The name of the possible variable which monitor the algorithm.
The frequency for a systematic drawing or the probability for a quantile drawing.
rwg
:A "numeric"
providing the variable weights to compute the distance determining the best representative to be determined when rty[1]
is '0' (their order is supposed to be this of slot nat
).
rkk
:A "numeric(1)"
: the power coefficient in the distance formulae (cannot be negative).; necessary only when slot rwg
exists.
see print8win
for the details.
J.-B. Denis
1 | showClass("win")
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.