Built-in tests
Independent Binomials
EqualitySampler.Simulations.proportion_test
— Functionproportion_test(
successes::AbstractArray{T<:Integer, 1},
observations::AbstractArray{T<:Integer, 1},
partition_prior::Union{Nothing, AbstractPartitionDistribution};
spl,
mcmc_settings,
ϵ,
n_leapfrog,
kwargs...
) -> Any
Fit independent binomials to the successes and observations and explore equality constraints among the probabilities.
Arguments
successes
, vector of successes.observations
vector of no trials.partition_prior
, the prior to use over partitions ornothing
, which implies sampling from the full model.
Keyword arguments
spl
, overwrite the sampling algorithm passed to Turing. It's best to look at the source code for the parameter names and so on.mcmc_settings
, settings for sampling.ϵ
, passed toTuring.HMC
, only used whenpartition_prior !== nothing
.n_leapfrog
, passed toTuring.HMC
, only used whenpartition_prior !== nothing
.kwargs...
, passed toAbstractMCMC.sample
.
Post Hoc Tests in One-Way ANOVA
EqualitySampler.Simulations.anova_test
— Functionanova_test(
f::StatsModels.FormulaTerm,
df::DataFrames.DataFrame,
args...;
kwargs...
) -> Any
Using the formula f
and data frame df
fit a one-way ANOVA.
anova_test(
y::AbstractVector{<:AbstractFloat},
g::AbstractVector{<:Integer},
args...;
kwargs...
) -> Any
Using the vector y
and grouping variable g
fit a one-way ANOVA.
anova_test(
y::AbstractVector{<:AbstractFloat},
g::AbstractVector{<:UnitRange{<:Integer}},
args...;
kwargs...
) -> Any
Using the vector y
and grouping variable g
fit a one-way ANOVA. Here g
is a vector of UnitRanges where each element indicates the group membership of y
.
anova_test(
df::Union{DataFrames.DataFrame, EqualitySampler.Simulations.SimpleDataSet},
partition_prior::Union{Nothing, AbstractPartitionDistribution};
spl,
mcmc_settings,
modeltype,
rng
) -> Any
Arguments:
df
a DataFrame or SimpleDataSet.partition_prior::Union{Nothing, AbstractPartitionDistribution}
, either nothing (i.e., fit the full model) or a subtype ofAbstractPartitionDistribution
.
Keyword arguments
spl
, overwrite the sampling algorithm passed to Turing. It's best to look at the source code for the parameter names and so on.mcmc_settings
, settings for sampling.modeltype
,:old
indicated all parameters are sampled whereasreduced
indicates onlyg
and the partitions are sampled using an integrated representation of the posterior.rng
a random number generator.