===========
 Data Sets
===========

.. _data-parties-positions:

Data on Parties' Positions in Multiple Dimensions across Time and Space
=========================================================================

|zenodo|_

.. grid:: 1 2 2 2
               
   .. grid-item::

      .. figure:: data/econlr.png

          Locations of objectives and parties' positions in the economic policy space

   .. grid-item::
   
      .. figure:: data/nonmat-objectives.png
   
         Locations of parties' positions in the non-material policy space

This page provides access to data on parties' positions in latent
political spaces, reconstructed based on data from the `Manifesto
Project <https://manifestoproject.wzb.eu/>`__ (also known as the
'Comparative Manifestos Project') using the method published in
*Political Analysis* as `"A Dynamic State-Space Model of Coded Political
Texts" </research/publications/dynamic-model-coded-texts/>`__. While there is a
'data verse' repository associated with that publication, I make the
data available at this site because that way it is much easier to
maintain and update. In particular you will find here parties' reconstructed positions on

1. an economic left/right dimension (or interventionist/free-market
   dimension) (in variable ``econlr``)
2. an authoritarian/libertarian dimension (in variable ``authlib``)
3. an traditionalist/modern dimension (in variable ``tradmod``)

The data come in two variants:

1. `point predictions with measures of
   uncertainty </data/party-positions/partypos-summaries.csv>`__
   (more specifically: posterior means and variances and 95%
   posterior prediction intervals)
2. multiple imputation values for more careful analysis of party
   positions as dependent or independent variables. There are two
   versions currently available: with `5 replications for each party
   and each time
   point </data/party-positions/partypos-imputations-5.csv>`__,
   with `100 replications for each party time
   point </data/party-positions/partypos-imputations-100.csv.zip>`__,
   and with `2000 replications for each party time
   point </data/party-positions/partypos-imputations-100.csv.zip>`__.
   The 2000-replications file is quite large, so please download it
   only if you know that you will need it.

The parties in the corresponding data sets are identified by their
IDs as in the Manifesto Project data sets.

The current version of the software (an R-package) along with documentation 
is available `here </software/manifestos/>`_.

If you use the party position data available here, please
cite the `Manifesto Project <https://manifestoproject.wzb.eu/>`__ in the
way they request *and* cite the `Political Analysis
article </research/publications/dynamic-model-coded-texts/>`__ as describing the
way the party positions were reconstructed, as well as the `DOI given by zenodo`_.

If you use the software,
please cite the `Political Analysis
article </research/publications/dynamic-model-coded-texts/>`__ as well as the
"manifestos" R package.

.. |zenodo|  image:: data/zenodo.3829630.svg
.. _zenodo: https://doi.org/10.5281/zenodo.3829630
.. _DOI given by zenodo: https://doi.org/10.5281/zenodo.3829630

