tempun

Citation

Vojtěch Kaše. (2022). tempun (Version v0.2). Zenodo. http://doi.org/10.5281/zenodo.4650989

Description

tempun is a Python 3 package to deal with temporal uncertainty in historical & archaeological datasets. Dating of historical artifacts (e.g. inscriptions, books etc.) is typically expressed by means of a dating interval during which it is assumed that the artifact was produced. Commonly, the production itself was much shorter than the interval and did not take more than one year. Therefore, the interval rather expresses uncertainty concerning the actual date of production. The question is how to analyze temporal trends in data revealing such temporal uncertainty.

A way forward is to use the interval to extract probability of production of the artifact. All dates (years) outside of the dating interval have probality p=0; all dates within the interval have probability somehow proportional to the duration of the interval, while the sum of probabilities for all years within the interval has to be equal to 1. This probality has to follow certain distribution. The package works with uniform or trapezoidal distribution. With uniform distribution, each year within the interval has an equal probality to be the year of production of the artifact.

We can use the intervals and the probalities associatiated with them to randomly assign individual dates (years) to each artifact within our dataset. In other words, we can model or simulate the date. We can do this repeatedly, i.e. to each artifact assign a certain number of random dates. This is in the core of a Monte Carlo Simulation (MCS) approach. In the package, it is implemented by means of model_date() function.

Having the random dates, we can proceed to do the analysis. For instance, we can recombine these dates into multiple time series and to compare between them. The package includes a bunch of functions developed for this purpose.

Getting started

The package can be installed via pip:

pip install tempun

To be sure that you have the latest version, use pip install tempun --ignore-installed. To install it directly from Jupyter, use !pip install tempun.

In Python, import the package:

import tempun

## Documentation (in progress)

model_date()

This function requires at least two parameters:

• start
• stop

If both start and stop are numbers, model_date(start, stop) returns a random number within the range starting with start and ending with stop.

If stop is not a valid number or contains an empty value, start is interpreted as defining a NOT BEFORE date (the so called ante quem*)

If start is not a valid number or contains an empty value, stop is interpreted as defining a NOT AFTER date (the so called post quem)

If start and stop are identical, the function returns the same number as well.

There are three optional parameters:

• size=1: how many random numbers you want to get; by default, size=1, i.e. only one number is returned
• b: bending point b defining shape of the trapezoidal distribution; by default, b=0, i.e. uniform distribution; set to 0.1 to get trapezoidal distribution
• scale: scale of the half-uniform distribution used to model ante quem and post quem; by default scale=25

The function returns an individual number (if size=1; i.e. by default) or a list of numbers of length equal to size

# example 1: only start and stop
>>> tempun.model_date(-340, -330)
-337
# example 2: size specified (returns a list of numbers of given size
>>> tempun.model_date(-340, -330, 10)
[-334, -333, -332, -336, -332, -338, -333, -336, -333, -331]
# example 3: model post quem (with default scale)
>>> tempun.model_date(114, "", 10, antepost=False)
[114, 114, 114, 114, 114, 114, 114, 114, 114, 114]
>>> tempun.model_date(114, "", 10, antepost=True)
[123, 143, 123, 149, 123, 155, 125, 115, 128, 132]

get_simulation_variants()

>>> get_simulation_variants(random_dates_lists)

timeblocksplot_from_random()

>>> timeblocksplot_from_randoms(random_dates_lists, timeblocks=None, ax=None, color="black", random_size=None, **kwargs)

kdeplot_from_randoms()

>>>kdeplot_from_randoms(random_dates_lists, ax=None, color="black", random_size=None, **kwargs)

timeblocksplot_from_randoms()

>>>timeblocksplot_from_randoms(random_dates_lists, timeblocks=None, ax=None, color="black", random_size=None, **kwargs)

Version history

• 0.2.6 - solving dependencies
• 0.2.5 - fixed problems in model_date returning NoneType under some conditions
• 0.2.4 - import from scipy.stats import trapz replaced by from numpy import trapz
• 0.2.3 - import from scipy.stats import trapz replaced by from scipy.integrate import trapz
• 0.2.2 - minor bugs fixing
• 0.2.1 - minor bugs fixing
• 0.2 - numerous improvements and simplifications (some new features not compatible with previous versions...)
• 0.1.6 - minor improvements
• 0.1.5 - antepost argument added, default False
• 0.1.4 - seed argument in model date
• 0.1 - first version