About hip-analysis

Motivation

Analysis in support of humanitarian early warning and assistance rely on the joint analysis of data from different domains (e.g. climate, prices, food security) and coming in different formats (e.g. raster, points, administrative statistics). RAM has invested in data layer applications to easily serve such varied types of data. However, the downstream analysis of such data is done using a heterogeneity of tools, and becomes particularly difficult it involves such mix-mode data. Additionally, before hip-analysis, there is no structured library for RAM analyses in python.

hip-analysis is a general-purpose analysis package to help support RAM-C analysis use cases, and, aspirationally, those of the whole of RAM. The package is designed to offer easy to use, modular components to carry out routine analytical steps, like sourcing data, preprocessing it, transforming it and visualizing results. The package has two main goals: - To offer a toolbox of repeatable analyses relevant for humanitarian operations via few, intuitive, lines of code. This reduces the barrier to entry for analysts, particularly in the field, to access powerful analysis pipelines independently as well develop new ones. - To transparently document the analytical logic of each pipeline. This facilitates knowledge sharing, collaboration, reuse, improvement, and extension of the toolbox.

Design principles

The nature of the task of generating evidence for humanitarian operations has determined the following design principles:

Analyses are spatio-temporally contextualized. Humanitarian analysis are always situated in a geography and time. The package should allow a user can define their overall area of interest and time frame, as well as events and locations of particular interest. The area of interest could also be the whole globe.

Analyses must relate many varied data structures. These include raster, geometries, points, survey data, administrative statistics etc. (Nearly) all such data has a spatio-temporal dimension (even if imprecise and/or aggregated). As such, all data can be combined, or at least related, spatially. The package should provide tools to carry out this spatial-temporal relation easily.

Data should be easy to access. The package has a common, extensible, high-level API to connect to varied data sources, in particular: existing RAM data layers; key external data sources, and user custom datasets.

Analytical components should be reusable and composable. Humanitarian analyses often rely common and standard analytical steps. The package modularity will allows users to flexibly combine such analytical building blocks into custom end-to-end pipelines for bespoke use cases.

The barriers to using the package should be low. The most popular and routine analyses should be usable by the widest possible community of analysts. The package will contains a catalogue of repeatable analyses that can be triggered over a chosen area/time with only few lines of code.