Description Usage Arguments Details Value Model description Examples
Functions for storing, plotting and model construction for noise
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 | ## Default S3 method:
noise.model(x, trend,
offset = 0,
model = "estimate",
reg.type = c("winsor", "trim"),
reg.level = 0,
averaging.type = c("sliding-window", "quantile-break", "equal-break", "none"),
breaks = 2, window = 51,
FUN = median,
FUN.trend = FUN,
na.rm = TRUE,
...)
## S3 method for class 'BioSSA2d'
noise.model(x, groups, ...)
## S3 method for class 'noise.model'
plot(x,
absolute = FALSE, #TODO Mb, remove it????
relative = TRUE,
draw.residuals = TRUE,
draw.means.fitted = FALSE,
print.alpha = TRUE,
ref = TRUE,
symmetric = !absolute,
...,
dots.residuals = list(),
dots.means.fitted = list(),
digits = max(3, getOption("digits") - 3))
## S3 method for class 'noise.model'
summary(object, digits = max(3, getOption("digits") - 3), ...)
## S3 method for class 'noise.model'
print(x, digits = max(3, getOption("digits") - 3), ...)
|
x, object |
the input object. This might be ‘BioSSA2d’/‘BioSSA3d’ object for BioSSA2d/BioSSA3d method, or just a numeric vector of residuals for default implementation |
trend |
numeric vector, trend for noise model estimation; this parameter can be used only for default method |
offset |
numeric value, trend offset for noise model estimation |
model |
model name ('additive', 'multiplicative' or 'poisson') or multiplicity power value. Use 'estimate' for perform model estimation |
reg.type |
regularization type for residuals and trend values averaging |
reg.level |
numeric value, quantile level for regularization |
averaging.type |
character, averaging method for power estimation and model visualization |
breaks |
the number of intervals to trend values breaking. Used for 'equal-break' and 'quantile-break' averaging types |
window |
length of sliding window. Used for 'sliding-window' averaging type |
FUN |
averaging function for logarithms of residuals absolute values
(like |
FUN.trend |
averaging function for logarithms of trend values, like previous |
na.rm |
a logical value indicating whether 'NA' values should be stripped before the computation proceeds |
... |
additional arguments, passed to inner function calls |
groups |
numeric vector, component indices in BioSSA decomposition for trend extraction |
absolute |
logical, whether plot absolute values of residuals instead of residuals itself |
relative |
logical, whether plot relative residuals instead of absolute ones |
draw.residuals |
logical, whether plot residuals |
draw.means.fitted |
logical, whether plot regression line drawn by averaged residuals |
print.alpha |
logical, whether output multiplicity power estimation |
ref |
logical, whether plot zero level line |
symmetric |
logical, whether y-scale should be symmetric by default |
dots.residuals |
list of additional arguments passed to |
dots.means.fitted |
same as previous, but for regression line plot (not used, supposed to be excluded) |
digits |
integer, how many significant digits are to be used for numbers formatting |
noise.model
applied to ‘BioSSA2d’/‘BioSSA3d’ object calculates the residual and trend
and then call the default method passing the further arguments.
plot
method plots residuals and smoothed residuals in dependence on the trend.
Data consisting of trend and residual values are ordered by the trend values,
then FUN
function is applied to residuals and the result is depicted
again FUN.trend
applied to the ordered trend values.
Object of class ‘noise.model’ for noise.model
. Trellis object for plot
.
Original object (invisibly) for print
and summary
.
Object of ‘noise.model’ is a list with following fields:
is numeric value, which is estimated if model = 'estimate'
,
or is equal to the user specified value (multiplicity power)
is standard deviation of the relative noise, (see ‘Model description’)
is square root of mean square deviation of relative residuals
is vector of residuals
is vector of trend values
is vector of smoothed (averaged) residuals absolute values
is vector of fitted values (by lm
)
of smoothed residuals absolute values
is vector of smoothed trend values
is offset value
is character name of used averaging type
is object of class ‘call’, constructor call (MB, drop it? This is command, which has created object, ‘call’ means ‘function call’)
Generalized multiplicative noise model is considered:
res_i = σ (trend_i + offset)^α \cdot ξ_i,
where ξ_i have standard normal distribution.
Value of alpha is estimated as follows:
logarithms of absolute values of residuals and ofsetted trend are considered,
then they both ordered by the trend values and averaging procedure performed.
There are following averaging methods:
'none' means nothing averaging,
'sliding-window' means sliding window averaging (default approach)
with window length denoted by window
argument,
'equal-break' and 'quantile-break' mean splitting all trend values
into buckets ('quantile-break' means buckets with equal quantity of elements in each bucket and
'equal-break' means equal size buckets), correspondingly residuals splitting and averaging
logarithmed residuals and trend values in each bucket.
Then linear regression on averaged logarithms is provided. Slope is alpha estimation and intercept is estimation of logarithm of the standard deviation of relative noise, i.e. sigma = exp(Intercept).
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 | xlim <- c(22, 88)
ylim <- c(32, 68)
L <- c(15, 15)
file <- system.file("extdata/data", "ab16.txt", package = "BioSSA")
df <- read.emb.data(file)
bs <- BioSSA(cad ~ AP + DV, data = df, ylim = ylim, xlim = xlim, L = L)
nm <- noise.model(bs, 1:3, averaging.type = "none")
plot(nm)
summary(nm)
nm <- noise.model(bs, 1:3, averaging.type = "sliding")
plot(nm)
summary(nm)
nm <- noise.model(bs, 1:3, averaging.type = "equal")
plot(nm)
summary(nm)
nm <- noise.model(bs, 1:3, averaging.type = "quantile")
plot(nm)
summary(nm)
nm <- noise.model(bs, 1:3, model = "poisson")
plot(nm)
summary(nm)
nm <- noise.model(bs, 1:3, model = "additive")
plot(nm)
summary(nm)
nm <- noise.model(bs, 1:3, model = "multiplicative")
plot(nm)
summary(nm)
nm <- noise.model(bs, 1:3, model = -1.2)
plot(nm)
summary(nm)
nm.none <- noise.model(bs, 1:3, model = "estimate", averaging.type = "none")
nm <- noise.model(bs, 1:3, model = nm.none$alpha, averaging.type = "sliding")
plot(nm)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.