weak-instruments 0.1.2

A package for weak instruments in an instrumental variables regression.

Solutions to Weak Instruments

Two stage least squares with instruments has become a common method for estimation. Issues can arise when there are many instruments and is compounded when there are many weak instruments. Problems also arise when treatment is clustered. Many methods have been proposed to handle these issues. In this file, we highlight some proposed solutions to many instruments bias and weak instruments bias, and run Monte Carlo simulations on a newly created Python package for handling many instruments bias.

JIVE1 and JIVE2 estimates based on

UJIVE1 and UJIVE2 estimates based on Angrist, Imbens, and Krueger (1999)

We use the following formula for estimation of JIVE1:

$[\frac{Z_i \hat{\pi} - h_i X_i}{1-h_i}]$

where $h_i$ is the leverage for observation $i$.

We use the following formula for estimation of JIVE2:

$[\frac{Z_i \hat{\pi} - h_i X_i}{1-(\frac{1}{N})}]$

Cluster Jive (CJIVE)

IJIVE

SJIVE

LIML

Two Stage Least Squares

Weak identification with many instruments (Mikushueva and Sun)

Lim et al.

Jacknife Anderson-Rubin tests for many weak IV inference

Lagrange Multiplier

HFUL

Each file should check for:

• Multicollinearity
• Perfect collinearity
• Dimensions of variables (Single column of controls etc)
  • Check to see if dimensions are the same for all variables
• Constant columns
• Y and Z must be one dimensional vectors