Sediment Analysis Tools

Import, store, and visualize sediment data.

Grain(v::AbstractVector{T}, measurement_names::AbstractVector{String})

Struct to hold grain level data

DensityTensor(KDEs, domains, sinks)
DensityTensor(array, domains, measurement_names; kw...)

An order 3 array to hold the density distributions for multiple sinks.

KDEs is a Vector{Vector{Vector{T}}} like type whereas array is an Array{T,3} like type.

Call setsourcename! to set the source name (name of first dimention).

Alias for Sink

Sink(grain1, grain2, ...)
Sink([grain1, grain2, ...])

Collects a list of Grains into a Rock/Sink.

Ensures all Grains have the same names and are in the same order. Struct to hold sink level data

Alias for Sink

Base.getindex(s::Sink, k::String)
Base.getindex(s::Sink, ks::AbstractVector{String})

Gets all values of the measurement k in the Sink.

names(n::NamedArray, dimname::Union{String,Symbol})

Extend the names function from NamedArray to get the names given the axis name rather than the axis number.

Plots.heatmap(M::NamedMatrix; kwargs...)

Automatically grabs the names in axis to label the x and y plot axis

make_densities(s::Sink; kwargs...)
make_densities(s::Sink, domains::AbstractVector{<:AbstractVector}; kwargs...)

Estimates the densities for each measurement in a Sink.

When given domains, a list where each entry is a domain for a different measurement, resample the kernel on this domain.

Parameters

• bandwidths::AbstractVector{<:Real}: list of bandwidths used for each measurement's

density estimation

• inner_percentile::Integer=100: value between 0 and 100 that filters out each measurement

by using the inner percentile range. This can help remove outliers and focus in on where the bulk of the data is.

Returns

• density_estimates::Vector{UnivariateKDE}

make_densities2d(s::Sink; kwargs...)
make_densities2d(s::Sink, domains::AbstractVector{<:AbstractVector}; kwargs...)

Similar to make_densities but performs the KDE on 2 measurements jointly.

read_raw_data(filename)
read_raw_data(filename; skip_sheets)

Imports excel data to a Vector{Sink}.

Excel file must have one element per page where different columns correspond to different sinks. Each sink can have a different number of grains (length of the column), but a sink must have a consistant length across different measurements (sheets).

Optionaly provide a collection skip_sheets to blacklist sheet names from the excel file.

Turn a DensityTensor or NamedArray into a plain Array type.

getdomain(D::DensityTensor, measurement::String)
getdomain(D::DensityTensor, j::Integer)

Gets the domain for the measurement density, the locations where the density was sampled. See getdomains.

getdomains(D::DensityTensor{T})::Vector{Vector{T}}

Gets the domain for every measurement's density, the locations where each density was sampled. See getdomains.

getmeasurements(g::Grain)
getmeasurements(s::Sink)
getmeasurements(D::DensityTensor)

Getter for the measurement names.

getmeasurements(s::Sink)

Gets the names of measurements from a Sink

getsourcename(D::DensityTensor)

Gets the name for the grouping of measurements. Usually "Sink" or "Source". See getsourcename.

getsourcenames(D::DensityTensor)

Gets the list of all sources' names. For example, ["Sink 1", "Sink 2", ...]. See getsourcenames.

getstepsizes(D::DensityTensor)

Gets the step sizes used for each domain. See getdomains.

Turn a DensityTensor into a NamedArray type.

Alias for getsource

getsource(D::DensityTensor, i::Integer)
getsink(D::DensityTensor, i::Integer)

Gets source/sink i from D. See eachsource.

normalize_density_sums!(D::DensityTensor)

Rescales the densities so that the sum of the density samples is 1.

This is in constrast to the usualy normalization for density functions where the area of the density curve is 1. In the case of an evenly sampled density, this area is sum(density_samples)*step_size.

Use normalize_density_sums to avoid mutation.

See normalize_density_sums!

setsourcename!(D::DensityTensor, name::String)

Sets the name of the source used by getsourcename and getsourcenames.

match_sources!(C, F, C_true, F_true)

Permute sources in C and F to match the ground truth C_true and F_true.

Similarity is checked by finding the source that minimizes norm(c - c_true) where c is the column of C.

Parameters

• double_check::Bool: When true, repeat for F and F_true with their horizontal slices, and assert the ordering is the same as before.

confidence_score(source_likelihoods; sorted=false)

Scores each grain between [0, 1] on how confident it came from the most likely source.

The likelihoods should be sorted in desending order. If they are presorted, use the keyword sorted=true.

Input

source_likelihoods: A list-of-lists or NTuple of Vectors of the likelihood each grain came from each source.

estimate_which_source(grain::Grain, F::DensityTensor; kwargs...)

Returns the likelihood and source index of the mostly likely factor the grain vector came from.

Returns

• (Default) source_index::Integer: The index of the most likely source
• (when max_likelihoods==true) (maxlikelihood, source_index): The most likely source and its likelihood
• (when all_likelihoods==true) likelihoods::Vector{Real}: Likelihood grain came from each source
• (when both are true) ((maxlikelihood, source_index), likelihoods)

estimate_which_2d_source(grain::Grain, F; kwargs...)

Returns the likelihood and source index of the mostly likely factor the grain vector came from.

Works on densities F where each horizontal slice is a discretized 2D KDE.

See estimate_which_source.

Returns

• (Default) source_index::Integer: The index of the most likely source
• (when max_likelihoods==true) (maxlikelihood, source_index): The most likely source and its likelihood
• (when all_likelihoods==true) likelihoods::Vector{Real}: Likelihood grain came from each source
• (when both are true) ((maxlikelihood, source_index), likelihoods)

estimate_which_nd_source(grain::Grain, F; kwargs...)

Returns the likelihood and source index of the mostly likely factor the grain vector came from.

Works on densities F where each first order slice is a discretized n-Dimentional KDE.

See estimate_which_source and estimate_which_2d_source.

Returns

• (Default) source_index::Integer: The index of the most likely source
• (when max_likelihoods==true) (maxlikelihood, source_index): The most likely source and its likelihood
• (when all_likelihoods==true) likelihoods::Vector{Real}: Likelihood grain came from each source
• (when both are true) ((maxlikelihood, source_index), likelihoods)

label_accuracy(labels, true_amounts::AbstractArray{T,3})

Calculates the number of correctly labeled grains and percentage of correct labels

Arguments

• labels: Iterable of iterables with the labels for each grain (Vector{Vector{T}} or Tuple{Tuple})
• true_amounts: Number of grains from each source for every sink

Outputs

• n_correct_eachsink::Integer
• n_total_labels::Integer
• accuracy::Real

eachdensity(D::DensityTensor)
eachdensity(D::DensityTensor, measurement::String)
eachdensity(...; sink::Integer)

Iterates D over each density vector. These are the 3 fibers of D. If a measurement is given, iterates over the densities for that measurement. If the keyword sink is given, only show densities from that sink.

eachmeasurement(s::Sink)

Iterator for a list of values of each measurement

eachmeasurement(D::DensityTensor)

Iterates D over each measurement slice. These are the lateral slices.

Alias for eachsource.

eachsource(D::DensityTensor)
eachsink(D::DensityTensor)

Iterates D over each source/sink slice. These are the horizontal slices. See getsource.

measurement_heatmaps(D::DensityTensor; kw...)

Returns heatmaps for each measurement (lateral slices) of D.

distributions_plot(D::DensityTensor, measurement::String; kw...)

Returns one plot will all distributions for a given measurement (lateral slice) of D.

source_heatmaps(D::DensityTensor; title="", kw...)

Returns heatmaps for each source (horizontal slices) of D.

The source name/index for each source will be appended to title.

plot_convergence(rel_errors, norm_grad, dist_Ncone)

Returns 3 separate plots for the three convergence metrics on a log10-y scale.

plot_convergence((rel_errors, norm_grad, dist_Ncone))

plot_source_index(
indexes::AbstractVector{<:Integer},
loglikelihood_ratios::AbstractVector{<:Real};
kwargs...
)

Returns one scatter plot of dots (eachindex(indexes), indexes) with brighter colours corresponding to higher loglikelihood_ratios.