Я использую пакет R frmpd для оценки модели QMLcre. Структура моих данных следующая:
str(BaseHORmodel)
Classes ‘spec_tbl_df’, ‘tbl_df’, ‘tbl’ and 'data.frame': 7830 obs. of 11 variables:
$ PortionFunding : num 0.00445 0.00359 0.00367 0.00125 0.00244 ...
$ Alignment : num 1 1 1 1 1 1 1 0 0 0 ...
$ RepresentativeVoteShare: num 0 0.605 0 0.631 0 ...
$ Democrat : num 0 0 0 0 0 0 0 0 0 0 ...
$ MajorityParty : num 1 1 1 1 1 1 0 0 0 0 ...
$ MPSpillOver : num 0 0 1 1 1 1 0 0 0 0 ...
$ PartyLeader : num 0 0 0 0 0 0 0 0 0 0 ...
$ Approriations : num 1 1 0 0 0 0 1 1 1 1 ...
$ PresidentVoteShare : num 0 0 0 0.625 0 ...
$ StateCD : chr "ALABAMA 1" "ALABAMA 1" "ALABAMA 1" "ALABAMA 1" ...
$ Year : num 2001 2002 2003 2004 2005 ...
- attr(*, "spec")=
.. cols(
.. Congress = col_double(),
.. Year = col_double(),
.. StateCD = col_character(),
.. PortionFunding = col_double(),
.. Alignment = col_double(),
.. RepresentativeVoteShare = col_double(),
.. Democrat = col_double(),
.. MajorityParty = col_double(),
.. MPSpillOver = col_double(),
.. PartyLeader = col_double(),
.. Whip = col_double(),
.. Approriations = col_double(),
.. PresidentVoteShare = col_double(),
.. total_funding_CNCS = col_double(),
.. total_funding_USDA = col_double(),
.. total_funding_DOC = col_double(),
.. total_funding_DOD = col_double(),
.. total_funding_ED = col_double(),
.. total_funding_DOE = col_double(),
.. total_funding_HHS = col_double(),
.. total_funding_HUD = col_double(),
.. total_funding_DOJ = col_double(),
.. total_funding_DOI = col_double(),
.. total_funding_DOT = col_double(),
.. total_funding_EPA = col_double(),
.. total_funding_FEMA = col_double(),
.. total_funding_IMLS = col_double(),
.. total_funding_NASA = col_double(),
.. total_funding_NEA = col_double(),
.. total_funding_NEH = col_double(),
.. total_funding_NSF = col_double(),
.. total_funding_ARC = col_double(),
.. total_funding_DOL = col_double(),
.. total_funding_FMCS = col_double(),
.. total_funding_SBA = col_double(),
.. total_funding_NA = col_double(),
.. total_funding_NARA = col_double(),
.. total_funding_USIP = col_double(),
.. total_funding_USAID = col_double(),
.. total_funding_DOS = col_double(),
.. total_funding_USACE = col_double(),
.. total_funding_JUSFC = col_double(),
.. total_funding_NLRB = col_double(),
.. total_funding_NRC = col_double(),
.. total_funding_SSA = col_double(),
.. total_funding_TREAS = col_double(),
.. total_funding_DHS = col_double(),
.. total_funding_EAC = col_double(),
.. total_funding_VA = col_double(),
.. total_funding_STB = col_double(),
.. total_funding_DENALI = col_double(),
.. total_funding_DRA = col_double(),
.. total_funding_AFRH = col_double(),
.. total_funding_EOP = col_double(),
.. total_funding_NCUA = col_double(),
.. total_funding_GCERC = col_double(),
.. total_funding_BBG = col_double(),
.. total_funding_MCC = col_double()
.. )
Я использую следующий код
library(tidyverse)
library(frmpd) # fractional regression models
library(plm) # has FE regression
BaseHORmodel <- read_csv("HORData.csv")
StateCD <- as.matrix(BaseHORmodel$StateCD)
YearCD <- as.matrix(BaseHORmodel$Year)
Y <- as.matrix(BaseHORmodel$PortionFunding)
X <- as.matrix(BaseHORmodel[,2:9])
HORQMLcre <- frmpd(id = StateCD, time = YearCD,
y = Y, x = X,
type = "QMLcre", table = TRUE, link = "probit", var.type = "cluster",
tdummies = TRUE)
Однако он выдает такой вывод:
*** Fractional probit regression model ***
*** Estimator: QMLcre
*** Exogeneity: TRUE
*** Use first lag of instruments: FALSE
Estimate Std. Error t value Pr(>|t|)
Alignment 0.015775 . . .
RepresentativeVoteShare 0.203812 . . .
Democrat 0.075075 . . .
MajorityParty 0.021212 . . .
MPSpillOver -0.005518 . . .
PartyLeader 0.023212 . . .
Approriations 0.038570 . . .
PresidentVoteShare 1.059441 . . .
time.2002 -0.135448 . . .
time.2003 -0.001838 . . .
time.2004 -0.735754 . . .
time.2005 -0.006845 . . .
time.2006 -0.145231 . . .
time.2007 -0.016142 . . .
time.2008 -0.777101 . . .
time.2009 -0.025628 . . .
time.2010 -0.155120 . . .
time.2011 -0.005779 . . .
time.2012 -0.734481 . . .
time.2013 -0.004031 . . .
time.2014 -0.137728 . . .
time.2015 -0.002429 . . .
time.2016 -0.724418 . . .
time.2017 -0.000792 . . .
time.2018 -0.133673 . . .
INTERCEPT_7 2.358415 . . .
INTERCEPT_11 -0.510523 . . .
INTERCEPT_17 -2.090756 . . .
INTERCEPT_18 -3.600679 . . .
Alignment_mean_7 1.150922 . . .
Alignment_mean_11 1.360588 . . .
Alignment_mean_17 1.375427 . . .
Alignment_mean_18 -0.000317 . . .
RepresentativeVoteShare_mean_7 -1.127225 . . .
RepresentativeVoteShare_mean_11 -2.146366 . . .
RepresentativeVoteShare_mean_17 -5.493256 . . .
RepresentativeVoteShare_mean_18 -0.028574 . . .
Democrat_mean_7 0.199227 . . .
Democrat_mean_11 0.256399 . . .
Democrat_mean_17 0.001506 . . .
Democrat_mean_18 0.232011 . . .
MajorityParty_mean_7 NA . . .
MajorityParty_mean_11 -1.975176 . . .
MajorityParty_mean_17 -2.153824 . . .
MajorityParty_mean_18 -0.012146 . . .
MPSpillOver_mean_7 -1.471042 . . .
MPSpillOver_mean_11 0.755698 . . .
MPSpillOver_mean_17 -0.231129 . . .
MPSpillOver_mean_18 0.192625 . . .
PartyLeader_mean_7 NA . . .
PartyLeader_mean_11 NA . . .
PartyLeader_mean_17 NA . . .
PartyLeader_mean_18 0.233615 . . .
Approriations_mean_7 -0.835486 . . .
Approriations_mean_11 -1.067978 . . .
Approriations_mean_17 0.470128 . . .
Approriations_mean_18 0.005492 . . .
PresidentVoteShare_mean_7 -32.872939 . . .
PresidentVoteShare_mean_11 -21.152605 . . .
PresidentVoteShare_mean_17 10.851473 . . .
PresidentVoteShare_mean_18 3.828431 . . .
Note: cluster standard errors
Number of observations (initial): 7830
Number of observations (for estimation): 7818
Number of cross-sectional units (initial): 459
Number of cross-sectional units (for estimation): 447
Average number of time periods per cross-sectional unit (initial): 17.05882
Average number of time periods per cross-sectional unit (for estimation): 17.48993
I Я не уверен, что делать. Глядя на документацию по frmpd, он говорит, что если для таблицы задано значение true, статистика тестов будет отображаться, однако это не так. Мне нужно посмотреть статистику теста, чтобы оценить причинное влияние Alginemt на частичное финансирование. Любой совет помогает.