dfabianus / Experiments.jl

Experiment is a data type for the collection and analysis of experimental time series data.

Experiments

-- Still in early development --

When collecting time series experimental data, the frequency and completeness of the single measurements often differs a lot. Different measurements are usually not all taken at exactly the same time and in the same frequence making it complicated to handle them in one single DataFrame. Subsequent data analysis and calculation of additional variables then requires to handle the varying quality and quantity of the data through interpolation and other measures.

This package builts on top of TimeSeries.jl and DataFrames.jl and should provide following features.

• Stroing of multiple TimeSeries with different timestamps in an Experiment object.
• Apply calculation functions to infer new TimeSeries based on the available ones.
• The package handles interpolation between the measurement TimeSeries automatically.
• Process parameters are also stored as TimeSeries of length 2, which are the start and end timepoints of the experiment.
• You can combine several Experiment objects into an array of Experiments allowing for calculations based on multiple experiments.
• An interface for DataFrames.jl enables DataFrame based operations on the data.

Quckstart tutorial

using Experiments
using Dates

# Define the first example timeseries with frequency of 0.1 hours (151 points).
t1 = DateTime(2018,1,1,0,0,0) .+ Experiments.hours2duration.(0:0.1:15)
v1 = sin.(0:0.1:15)
ts1 = timeseries(:measurement1, t1, v1)
# 151×1 TimeSeries.TimeArray{Float64, 2, DateTime, Matrix{Float64}} 2018-01-01T00:00:00 to 2018-01-01T15:00:00

# Define a second example timeseries with frequency of 0.25 hours (61 points).
t2 = DateTime(2018,1,1,1,0,0) .+ Experiments.hours2duration.(0:0.25:15)
v2 = cos.(0:0.25:15)
ts2 = timeseries(:measurement2, t2, v2)
# 61×1 TimeSeries.TimeArray{Float64, 2, DateTime, Matrix{Float64}} 2018-01-01T01:00:00 to 2018-01-01T16:00:00

We created two instances of type TimeArray from TimeSeries.jl by using the timeseries function.

using Plots
scatter(ts1)
scatter!(ts2)

yields the following plot

It can be noticed, that the signals are 1 hour shifted and the sampling frequency differs.

# Create an experiment with both timeseries
exp1 = Experiment("Experiment1", ts1, ts2) #Experiment1 experiment with 2 timeseries

# Use the timeseries command to get a TimeArray containing a specific measurement.
timeseries(exp1, :measurement1) # 151×1 TimeSeries.TimeArray{Float64, 2, DateTime, Matrix{Float64}} 2018-01-01T00:00:00 to 2018-01-01T15:00:00
timeseries(exp1, :measurement2) # 61×1 TimeSeries.TimeArray{Float64, 2, DateTime, Matrix{Float64}} 2018-01-01T01:00:00 to 2018-01-01T16:00:00

# Use the timeseries command to get a TimeArray containing both measurements.
timeseries(exp1) # 183×2 TimeSeries.TimeArray{Float64, 2, DateTime, Matrix{Float64}} 2018-01-01T00:00:00 to 2018-01-01T16:00:00
# Values are NaN, when there is no value available for the measurement at that timepoint.