API

struct Center

Represents a centering scheme, akin to StatsModels.AbstractContrasts. Pass as value in Dict as hints to schema (or as contrasts kwarg for fit).

Examples

Can specify center value to use:

julia> schema((x=collect(1:10), ), Dict(:x => Center(5)))
StatsModels.Schema with 1 entry:
  x => center(x, 5)

You can use a function to compute the center value:

julia> schema((x=collect(1:10), ), Dict(:x => Center(median))) StatsModels.Schema with 1 entry: x => x(centered: 5.5)

Or center will be automatically computed if omitted:

julia> schema((x=collect(1:10), ), Dict(:x => Center()))
StatsModels.Schema with 1 entry:
  x => center(x, 5.5)

struct CenteredTerm{T,C} <: AbstractTerm

A lazily centered term. A wrapper around an T<:AbstractTerm which will produce centered values with modelcols by subtracting center from each element generated by the wrapped term with modelcols.

Fields

• term::T: The wrapped term.
• center::C: The center value subtracted from the resulting modelcols.

Examples

Directly construct with given center:

julia> d = (x=collect(1:10), );

julia> t = concrete_term(term(:x), d)
x(continuous)

julia> tc = CenteredTerm(t, 5)
x(centered: 5)

julia> hcat(modelcols(t + tc, d)...)
10×2 Matrix{Int64}:
  1  -4
  2  -3
  3  -2
  4  -1
  5   0
  6   1
  7   2
  8   3
  9   4
 10   5

Construct with lazy centering via Center

julia> tc = concrete_term(term(:x), d, Center())
center(x, 5.5)

julia> hcat(modelcols(t + tc, d)...)
10×2 Matrix{Float64}:
  1.0  -4.5
  2.0  -3.5
  3.0  -2.5
  4.0  -1.5
  5.0  -0.5
  6.0   0.5
  7.0   1.5
  8.0   2.5
  9.0   3.5
 10.0   4.5

Or similarly via schema hints:

julia> sch = schema(d, Dict(:x => Center()))
StatsModels.Schema with 1 entry:
  x => center(x, 5.5)

struct Scale

Represents a scaling scheme, akin to StatsModels.AbstractContrasts. Pass as value in Dict as hints to schema (or as contrasts kwarg for fit).

Examples

Can specify scale value to use:

julia> schema((x=collect(1:10), ), Dict(:x => Scale(5)))
StatsModels.Schema with 1 entry:
  x => x(scaled: 5))

You can use a function to compute the scale value:

julia> schema((x=collect(1:10), ), Dict(:x => Scale(mad))) StatsModels.Schema with 1 entry: x => x(scaled: 3.71)

Or scale will be automatically computed if left out:

julia> schema((x=collect(1:10), ), Dict(:x => Scale()))
StatsModels.Schema with 1 entry:
  x => x(scaled: 3.03)

struct ScaledTerm{T,S} <: AbstractTerm

A lazily scaled term. A wrapper around an T<:AbstractTerm which will produce scaled values with modelcols by dividing each element by scale.

Fields

• term::T: The wrapped term.
• scale::S: The scale value which the resulting modelcols are divided by.

Examples

Directly construct with given scale:

julia> d = (x=collect(1:10), );

julia> t = concrete_term(term(:x), d)
x(continuous)

julia> ts = ScaledTerm(t, 5)
x(scaled: 5)

julia> hcat(modelcols(t + ts, d)...)
10×2 Matrix{Float64}:
  1.0  0.2
  2.0  0.4
  3.0  0.6
  4.0  0.8
  5.0  1.0
  6.0  1.2
  7.0  1.4
  8.0  1.6
  9.0  1.8
 10.0  2.0

Construct with lazy scaling via Scale

julia> ts = concrete_term(term(:x), d, Scale())
x(scaled: 3.03)

julia> hcat(modelcols(t + ts, d)...)
10×2 Matrix{Float64}:
  1.0  0.330289
  2.0  0.660578
  3.0  0.990867
  4.0  1.32116
  5.0  1.65145
  6.0  1.98173
  7.0  2.31202
  8.0  2.64231
  9.0  2.9726
 10.0  3.30289

Or similarly via schema hints:

julia> sch = schema(d, Dict(:x => Scale()))
StatsModels.Schema with 1 entry:
  x => scale(x, 3.03)

struct ZScore

Represents a z-scoring scheme, akin to StatsModels.AbstractContrasts. Pass as value in Dict as hints to schema (or as contrasts kwarg for fit).

Examples

Can specify the center and scale values to use:

julia> schema((x=collect(1:10), ), Dict(:x => ZScore(; center=5, scale=3)))
StatsModels.Schema with 1 entry:
  x => x(centered: 5 scaled: 3)

Or scale will be automatically computed if left out:

julia> schema((x=collect(1:10), ), Dict(:x => ZScore()))
StatsModels.Schema with 1 entry:
  x => x(centered: 5.5 scaled: 3.03)

struct ZScoredTerm{T,C,S} <: AbstractTerm

A lazily z-scored term. A wrapper around an T<:AbstractTerm which will produce scaled values with modelcols by subtracting center from each element and then dividing by scale.

Fields

• term::T: The wrapped term.
• center::C: The center value which is subtracted from the resulting modelcols.
• scale::S: The scale value which the resulting modelcols are divided by.

Examples

Directly construct with given scale:

julia> d = (x=collect(1:10), );

julia> t = concrete_term(term(:x), d)
x(continuous)

julia> ts = ZScoredTerm(t, 3, 5)
x(centered: 3 scaled: 5)

julia> hcat(modelcols(t + ts, d)...)
10×2 Matrix{Float64}:
  1.0  -0.4
  2.0  -0.2
  3.0   0.0
  4.0   0.2
  5.0   0.4
  6.0   0.6
  7.0   0.8
  8.0   1.0
  9.0   1.2
 10.0   1.4

Construct with lazy scaling via ZScore

julia> ts = concrete_term(term(:x), d, ZScore())
x(centered: 5.5 scaled: 3.03)

julia> hcat(modelcols(t + ts, d)...)
10×2 Matrix{Float64}:
  1.0  -1.4863
  2.0  -1.15601
  3.0  -0.825723
  4.0  -0.495434
  5.0  -0.165145
  6.0   0.165145
  7.0   0.495434
  8.0   0.825723
  9.0   1.15601
 10.0   1.4863

Or similarly via schema hints:

julia> sch = schema(d, Dict(:x => ZScore()))
StatsModels.Schema with 1 entry:
  x => x(centered: 5.5 scaled: 3.03)

_standard(xs::AbstractArray, val)

Translate an abstract standardization value to a concrete one based on xs.

nothing and already concrete Number vals are passed through. Otherwise, val(xs) is returned.

center!(f=mean, x, y=f(skipmissing(x)))

Center an array x in place about a scalar y.

See also center

center(f=mean, x, y=f(skipmissing(x)))

Center an array x about a scalar y.

See also center!

scale(f=std, x, y=f(skipmissing(x)))

Scale an array x in place by a scalar y.

See also scale

scale(f=std, x, y=f(skipmissing(x)))

Scale an array x by a scalar y.

See also scale!