Comparison of conditional F-statistics

In OneSampleMR: One Sample Mendelian Randomization and Instrumental Variable Analyses

# Only run code chunks if Suggests packages available
suggestsavailable <-
  sapply(c("lfe", "haven"), requireNamespace, quietly = TRUE)
evalcond <- all(suggestsavailable)
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  eval = evalcond
)

library(haven)
library(ivreg)
library(lfe)
library(OneSampleMR)

Run fsw() on ivreg() model object

url <- "http://fmwww.bc.edu/ec-p/data/wooldridge/mroz.dta"
dat <- haven::read_dta(url)
mod <- ivreg(lwage ~ educ + exper | age + kidslt6 + kidsge6, data = dat)
summary(mod)
fsw(mod)

Comparison with F-statistic from lfe package

modst2 <- felm(lwage ~ 1 | 0 | (educ | exper ~ age + kidslt6 + kidsge6),
                 data = dat)
summary(modst2)
t(sapply(modst2$stage1$lhs, 
         function(lh) waldtest(modst2$stage1, 
                               ~ age | kidslt6 | kidsge6, lhs = lh)))
condfstat(modst2, quantiles = c(0.025, 0.975))

Comparison with output from ivreg2

Code run using Stata version 16.1.

use http://fmwww.bc.edu/ec-p/data/wooldridge/mroz.dta, clear

// ssc install ivreg2
ivreg2 lwage (educ exper = age kidslt6 kidsge6) if !missing(lwage, educ, exper, age, kidslt6, kidsge6), first

First-stage regressions
-----------------------


First-stage regression of educ:

Statistics consistent for homoskedasticity only
Number of obs =                    428
------------------------------------------------------------------------------
        educ |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |  -.0185412   .0163449    -1.13   0.257    -.0506683    .0135859
     kidslt6 |   .6984283   .2966854     2.35   0.019     .1152709    1.281586
     kidsge6 |   -.222821   .0906154    -2.46   0.014    -.4009324   -.0447096
       _cons |   13.64009   .7644499    17.84   0.000     12.13751    15.14268
------------------------------------------------------------------------------
F test of excluded instruments:
  F(  3,   424) =     4.47
  Prob > F      =   0.0042
Sanderson-Windmeijer multivariate F test of excluded instruments:
  F(  2,   424) =     6.69
  Prob > F      =   0.0014


First-stage regression of exper:

Statistics consistent for homoskedasticity only
Number of obs =                    428
------------------------------------------------------------------------------
       exper |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |   .3948754   .0496446     7.95   0.000     .2972953    .4924555
     kidslt6 |  -.7469412   .9011267    -0.83   0.408    -2.518173    1.024291
     kidsge6 |  -1.430306   .2752275    -5.20   0.000    -1.971286   -.8893254
       _cons |  -1.500019   2.321874    -0.65   0.519    -6.063837    3.063798
------------------------------------------------------------------------------
F test of excluded instruments:
  F(  3,   424) =    55.04
  Prob > F      =   0.0000
Sanderson-Windmeijer multivariate F test of excluded instruments:
  F(  2,   424) =    81.81
  Prob > F      =   0.0000



Summary results for first-stage regressions
-------------------------------------------

                                           (Underid)            (Weak id)
Variable     | F(  3,   424)  P-val | SW Chi-sq(  2) P-val | SW F(  2,   424)
educ         |       4.47    0.0042 |       13.51   0.0012 |        6.69
exper        |      55.04    0.0000 |      165.17   0.0000 |       81.81

Stock-Yogo weak ID F test critical values for single endogenous regressor:
                                    5% maximal IV relative bias    13.91
                                   10% maximal IV relative bias     9.08
                                   20% maximal IV relative bias     6.46
                                   30% maximal IV relative bias     5.39
                                   10% maximal IV size             22.30
                                   15% maximal IV size             12.83
                                   20% maximal IV size              9.54
                                   25% maximal IV size              7.80
Source: Stock-Yogo (2005).  Reproduced by permission.
NB: Critical values are for Sanderson-Windmeijer F statistic.

Underidentification test
Ho: matrix of reduced form coefficients has rank=K1-1 (underidentified)
Ha: matrix has rank=K1 (identified)
Anderson canon. corr. LM statistic       Chi-sq(2)=13.10    P-val=0.0014

Weak identification test
Ho: equation is weakly identified
Cragg-Donald Wald F statistic                                       4.46

Stock-Yogo weak ID test critical values for K1=2 and L1=3:
                                   10% maximal IV size             13.43
                                   15% maximal IV size              8.18
                                   20% maximal IV size              6.40
                                   25% maximal IV size              5.45
Source: Stock-Yogo (2005).  Reproduced by permission.

Weak-instrument-robust inference
Tests of joint significance of endogenous regressors B1 in main equation
Ho: B1=0 and orthogonality conditions are valid
Anderson-Rubin Wald test           F(3,424)=       2.08     P-val=0.1025
Anderson-Rubin Wald test           Chi-sq(3)=      6.29     P-val=0.0983
Stock-Wright LM S statistic        Chi-sq(3)=      6.20     P-val=0.1023

Number of observations               N  =        428
Number of regressors                 K  =          3
Number of endogenous regressors      K1 =          2
Number of instruments                L  =          4
Number of excluded instruments       L1 =          3

IV (2SLS) estimation
--------------------

Estimates efficient for homoskedasticity only
Statistics consistent for homoskedasticity only

                                                      Number of obs =      428
                                                      F(  2,   425) =     3.03
                                                      Prob > F      =   0.0492
Total (centered) SS     =  223.3274513                Centered R2   =   0.1482
Total (uncentered) SS   =   829.594813                Uncentered R2 =   0.7707
Residual SS             =  190.2315236                Root MSE      =    .6667

------------------------------------------------------------------------------
       lwage |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        educ |   .1058361   .0806975     1.31   0.190    -.0523281    .2640003
       exper |   .0161527    .007568     2.13   0.033     .0013197    .0309858
       _cons |  -.3601821   1.029787    -0.35   0.727    -2.378528    1.658164
------------------------------------------------------------------------------
Underidentification test (Anderson canon. corr. LM statistic):          13.101
                                                   Chi-sq(2) P-val =    0.0014
------------------------------------------------------------------------------
Weak identification test (Cragg-Donald Wald F statistic):                4.463
Stock-Yogo weak ID test critical values: 10% maximal IV size             13.43
                                         15% maximal IV size              8.18
                                         20% maximal IV size              6.40
                                         25% maximal IV size              5.45
Source: Stock-Yogo (2005).  Reproduced by permission.
------------------------------------------------------------------------------
Sargan statistic (overidentification test of all instruments):           1.168
                                                   Chi-sq(1) P-val =    0.2798
------------------------------------------------------------------------------
Instrumented:         educ exper
Excluded instruments: age kidslt6 kidsge6
------------------------------------------------------------------------------

OneSampleMR documentation built on June 27, 2024, 1:06 a.m.