julia - 如何在 Julia 表达式中使用混合模型的 "|"给定符号

Question

我想以编程方式迭代一些因变量并使用 MixedModels 包创建一些公式，例如:

form = @formula(y ~ 1 + x1 + x2 + (1 + x1 + x2 | stuff))

理想情况下，我会有一个因变量符号数组，并迭代它们并创建不同类型的公式:

depvars = [:height, :weight]

for var in depvars
    my_formula = @formula(var ~ otherstuff + (1 + otherstuff | thing))
    lm_out = fit!(lmm(my_formula, data))
end

其中otherstuff和thing会发生变化。如果我没有混合模型，那么我可能会这样做:

depvars = [:height, :weight]
f1 = Expr(:call, :+, [:x1, :x2])
f2 = Expr(:call, :+, [:x3, :x4])

for var in depvars
    my_formula1 = Formula(var, f1)
    my_formula2 = Formula(var, f2)
    lm_out = lm(my_formula1, data)
    # and so on...
end

但我不确定是否可以在表达式中表达部分 (1 + x1 + x2 | stuff) 。

这可能吗？

Answer 1

不要直接使用@formula，而是使用DataFrames.Formula并在引用表达式中进行插值:

using MixedModels, DataFrames

depvars = [:height, :weight]
things = [:thing1, :thing2]
otherstuffs = [:other1, :other2]

for thing in things, otherstuff in otherstuffs
    for var in depvars
        my_formula = Formula(var, :( $otherstuff + (1 +$otherstuff | $thing)))
        @show my_formula
    end
end

给予:

my_formula = Formula: height ~ other1 + ((1 + other1) | thing1)
my_formula = Formula: weight ~ other1 + ((1 + other1) | thing1)
my_formula = Formula: height ~ other2 + ((1 + other2) | thing1)
my_formula = Formula: weight ~ other2 + ((1 + other2) | thing1)
my_formula = Formula: height ~ other1 + ((1 + other1) | thing2)
my_formula = Formula: weight ~ other1 + ((1 + other1) | thing2)
my_formula = Formula: height ~ other2 + ((1 + other2) | thing2)
my_formula = Formula: weight ~ other2 + ((1 + other2) | thing2)