You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
**MIDASpy** is a Python package for multiply imputing missing data using
28
+
deep learning methods. The **MIDASpy** algorithm offers significant
29
+
accuracy and efficiency advantages over other multiple imputation
30
+
strategies, particularly when applied to large datasets with complex
31
+
features. In addition to implementing the algorithm, the package contains
32
+
functions for processing data before and after model training, running
33
+
imputation model diagnostics, generating multiple completed datasets,
34
+
and estimating regression models on these datasets.
35
+
36
+
For an implementation in R, see our **rMIDAS** repository
37
+
[here](https://github.com/MIDASverse/rMIDAS).
38
+
39
+
## Background and suggested citation
40
+
41
+
For more information on MIDAS, the method underlying the software, see:
42
+
43
+
Lall, Ranjit, and Thomas Robinson. 2022. "The MIDAS Touch: Accurate and Scalable Missing-Data Imputation with Deep Learning." _Political Analysis_ 30, no. 2: 179-196. https://doi.org/10.1017/pan.2020.49.
44
+
45
+
An ungated version of the paper is available [here](http://eprints.lse.ac.uk/108170/1/Lall_Robinson_PA_Forthcoming.pdf).
46
+
47
+
## Installation
48
+
49
+
To install via pip, enter the following command into the terminal:
50
+
`pip install MIDASpy`
51
+
52
+
The latest development version (potentially unstable) can be installed
Interested in contributing to **MIDASpy**? We are looking to hire a research assistant to work part-time (flexibly) to help us build out new features and integrate our software with existing machine learning pipelines. You would be paid the standard research assistant rate at the London School of Economics. To apply, please send your CV (or a summary of relevant skills/experience) to [email protected].
79
+
80
+
## Version 1.2.3 (December 2022)
81
+
82
+
*v1.2.3 adds support for installation on Apple Silicon hardware (i.e. M1 and M2 Macs).*
83
+
84
+
## Version 1.2.2 (July 2022)
85
+
86
+
*v1.2.2 makes minor efficiency changes to the codebase. Full details are available in the Release logs.*
87
+
88
+
## Version 1.2.1 (January 2021)
89
+
90
+
*v1.2.1 adds new pre-processing functionality and a multiple imputation regression function.*
91
+
92
+
Users can now automatically preprocess binary and categorical columns prior to running the MIDAS algorithm using `binary_conv()` and `cat_conv()`.
93
+
94
+
The new `combine()` function allows users to run regression analysis across the complete data, following Rubin’s combination rules.
95
+
96
+
## Previous versions
97
+
98
+
*Version 1.1.1 (October 2020)*
99
+
100
+
Key changes:
101
+
102
+
- Update adds **full Tensorflow 2.X support**:
103
+
104
+
- Users can now run the MIDAS algorithm in TensorFlow 2.X (TF1 support
105
+
retained)
106
+
107
+
- Tidier handling of random seed setting across both TensorFlow and
108
+
NumPy
109
+
110
+
- Fixes a minor dependency bug
111
+
112
+
- Other minor bug fixes
113
+
114
+
*Version 1.0.2 (September 2020)*
115
+
116
+
Key changes:
117
+
118
+
- Minor, mainly cosmetic, changes to the underlying source code.
119
+
- Renamed ‘categorical\_columns’ argument in build\_model() to
120
+
‘binary\_columns’ to avoid confusion
121
+
- Added plotting arguments to overimputation() method to suppress
122
+
intermediary overimputation plots (plot\_main) and all plots
123
+
(skip\_plot).
124
+
- Changed overimputation() plot titles, labels and legends
125
+
- Added tensorflow 2.0 version check on import
126
+
- Fixed seed-setting bug in earlier versions
127
+
128
+
*Alpha 0.2:*
129
+
130
+
Variational autoencoder enabled. More flexibility in model
131
+
specification, although defaulting to a simple mirrored system. Deeper
132
+
analysis tools within .overimpute() for checking fit on continuous
<!-- [](https://codecov.io/gh/tidyverse/dplyr?branch=master) -->
**MIDASpy** is a Python package for multiply imputing missing data using
@@ -70,21 +56,21 @@ For a simple demonstration of **MIDASpy**, see our Jupyter Notebook
70
56
71
57
Interested in contributing to **MIDASpy**? We are looking to hire a research assistant to work part-time (flexibly) to help us build out new features and integrate our software with existing machine learning pipelines. You would be paid the standard research assistant rate at the London School of Economics. To apply, please send your CV (or a summary of relevant skills/experience) to [email protected].
72
58
59
+
## Version 1.2.3 (December 2022)
60
+
61
+
*v1.2.3 adds support for installation on Apple Silicon hardware (i.e. M1 and M2 Macs).*
62
+
73
63
## Version 1.2.2 (July 2022)
74
64
75
65
*v1.2.2 makes minor efficiency changes to the codebase. Full details are available in the Release logs.*
76
66
77
67
## Version 1.2.1 (January 2021)
78
68
79
-
*v1.2.1 adds new pre-processing functionality and a multiple imputation
80
-
regression function.*
69
+
*v1.2.1 adds new pre-processing functionality and a multiple imputation regression function.*
81
70
82
-
Users can now automatically preprocess binary and categorical columns
83
-
prior to running the MIDAS algorithm using `binary_conv()` and
84
-
`cat_conv()`.
71
+
Users can now automatically preprocess binary and categorical columns prior to running the MIDAS algorithm using `binary_conv()` and `cat_conv()`.
85
72
86
-
The new `combine()` function allows users to run regression analysis
87
-
across the complete data, following Rubin’s combination rules.
73
+
The new `combine()` function allows users to run regression analysis across the complete data, following Rubin’s combination rules.
0 commit comments