Fixed issue#2 and some warnings from CRAN

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: lfe
 Version: 2.8-1
-Date: 2018-08-20
+Date: 2018-08-22
 Title: Linear Group Fixed Effects
 Authors@R: person("Simen", "Gaure", email="[email protected]", role=c("aut","cre"))
 Copyright: 2011-2018, Simen Gaure
@@ -19,4 +19,5 @@ Classification/JEL: C13, C23, C60
 Classification/MSC: 62J05, 65F10, 65F50
 URL: https://github.com/sgaure/lfe
 BugReports: https://github.com/sgaure/lfe/issues
+Encoding: UTF-8
 RoxygenNote: 6.1.0

R/nlexpect.R

@@ -38,7 +38,7 @@
 #' one already.  It will also be passed an argument \code{.z} which contains
 #' the actual coefficients in normalized coordinates, i.e. if \code{ch} is the
 #' Cholesky decomposition of the covariance matrix, and \code{pt} are the point
-#' estimates, the coefficients will be \code{pt + ch \%*\% .s}. The first argument
+#' estimates, the coefficients will be \code{pt + ch \%*\% .z}. The first argument
 #' is a vector with names corresponding to the coefficients.
 #'
 #' If you specify \code{vectorized=TRUE}, your function will be passed a list with vectors
@@ -57,8 +57,8 @@
 #' The \code{...} can be used for passing other arguments to the integration
 #' routine.
 #' 
-#' @param est object of class \code{"felm"}, a result of a call to
-#' \code{\link{felm}}.
+#' @param est object of class \code{"felm"} or \code{"lm"}, a result of a call to
+#' \code{\link{felm}} or \code{lm}.
 #' @param fun function of coefficients to be integrated. Can also be a
 #' \code{quote}d expression.
 #' @param coefs character. Names of coefficients to test. Only needed if
@@ -111,13 +111,13 @@
 #' 2*nlexpect(est, x1*sign(pt1) < 0)
 #'
 #' # Here's a multivalued, vectorized example
-#' nlexpect(est, rbind(a=x1*x2 < pt1, b=x1*x2 > 0), vectorized=TRUE)
+#' nlexpect(est, rbind(a=x1*x2 < pt1, b=x1*x2 > 0), vectorize=TRUE)
 #' \donttest{
 #' 
 #' # Non-linear test:
 #'
 #' # A simple one, what's the probability that product x1*x2 is between 0 and |E(x1)|?
-#' nlexpect(est, x1*x2 > 0 & x1*x2 < abs(pt1), vectorized=TRUE)
+#' nlexpect(est, x1*x2 > 0 & x1*x2 < abs(pt1), vectorize=TRUE)
 #' # Then a more complicated one with the expected value of a polynomal in the coefficients
 #' f <- function(x) c(poly=x[['x1']]*(6*x[['x1']]-x[['x2']]^2))
 #' # This is the linearized test:

cleanup

@@ -1,4 +1,4 @@
 #!/bin/bash
-rm -f src/Makevars acaux config.log
+rm -rf src/Makevars src/config.h acaux config.* aclocal.m4 autom4te.cache
 
 

inst/NEWS.Rd

@@ -1,5 +1,13 @@
 \name{NEWS}
 \title{lfe news}
+\section{Changes in version 2.8-1}{
+  \itemize{
+    \item Fixed a bug in Crowsum which prevented computation of clustered standard
+    errors in large datasets (rows x covariates exceeds 2
+    billion). Thanks to mcmalone for the report.
+    \item Minor updates to documentation and some cosmetic code changes.
+  }
+}
 \section{Changes in version 2.8}{
   \itemize{
     \item Added functionality for closed form standard errors on the fixed

src/demean.c

@@ -279,7 +279,7 @@ static void *demeanlist_thr(void *varg) {
 #ifdef HAVE_THREADNAME
 #ifndef USEOMP
     char thrname[16];
-    snprintf(thrname,16, "Ct %d/%d",vecnum+1, arg->K);
+    snprintf(thrname,16, "Ct %5d/%5d",vecnum+1, arg->K);
     STNAME(thrname);
 #endif
 #endif

src/kaczmarz.c

@@ -217,7 +217,7 @@ static void *kaczmarz_thr(void *varg) {
     if(vecnum >= arg->numvec) break;
 #ifdef HAVE_THREADNAME
     char thrname[16];
-    snprintf(thrname, 16, "Kz %d/%d",vecnum+1, arg->numvec);
+    snprintf(thrname, 16, "Kz %5d/%5d",vecnum+1, arg->numvec);
     STNAME(thrname);
 #endif
 

tests/Examples/lfe-Ex.Rout.save

@@ -1,5 +1,5 @@
 
-R Under development (unstable) (2018-06-19 r74917) -- "Unsuffered Consequences"
+R Under development (unstable) (2018-08-19 r75170) -- "Unsuffered Consequences"
 Copyright (C) 2018 The R Foundation for Statistical Computing
 Platform: x86_64-pc-linux-gnu (64-bit)
 
@@ -59,9 +59,9 @@ Loading required package: Matrix
 > # find bias corrections
 > bccorr(est)
         corr           v1           v2          cov           d1           d2 
--0.032262721  0.952892562  1.376031164 -0.036943420  0.088877778  0.082733214 
+-0.030728526  0.950001416  1.370274827 -0.035059660  0.091768924  0.088489551 
          d12 
--0.004461406 
+-0.006345166 
 > 
 > 
 > 
@@ -434,8 +434,8 @@ function (v, addnames)
     }
     v
 }
-<bytecode: 0x556e19773ee0>
-<environment: 0x556e1772d320>
+<bytecode: 0x55f222b93e30>
+<environment: 0x55f2216d9148>
 > is.estimable(ef,fl)
 Warning in is.estimable(ef, fl) :
   non-estimable function, largest error 0.006 in coordinate 4013 ("id.4013")
@@ -457,8 +457,8 @@ function (v, addnames)
     }
     v
 }
-<bytecode: 0x556e18fe2758>
-<environment: 0x556e19253d08>
+<bytecode: 0x55f2224504f8>
+<environment: 0x55f22266dc28>
 attr(,"verified")
 [1] TRUE
 > is.estimable(ef,fl)
@@ -625,21 +625,21 @@ attr(,"df1")
 > # find bias corrections, there's little bias in this example
 > print(yv <- fevcov(est, lhs='y'))
               id         firm          foo
-id    0.92015755 -0.011173022 -0.019789538
-firm -0.01117302  0.963808718 -0.009781304
-foo  -0.01978954 -0.009781304  1.035240984
+id    0.92025530 -0.011069589 -0.019794399
+firm -0.01106959  0.964082331 -0.009815829
+foo  -0.01979440 -0.009815829  1.035318331
 attr(,"bias")
                 id          firm           foo
-id    0.0319481064 -2.052376e-04 -1.023073e-04
-firm -0.0002052376  2.187053e-02 -4.728025e-05
-foo  -0.0001023073 -4.728025e-05  1.560268e-02
+id    3.185035e-02 -3.086703e-04 -9.744607e-05
+firm -3.086703e-04  2.159692e-02 -1.275526e-05
+foo  -9.744607e-05 -1.275526e-05  1.552533e-02
 > ## Here's how to compute the unbiased correlation matrix:
 > cm <- cov2cor(yv)
 > structure(cm,bias=NULL)
               id         firm          foo
-id    1.00000000 -0.011864355 -0.020276072
-firm -0.01186435  1.000000000 -0.009792208
-foo  -0.02027607 -0.009792208  1.000000000
+id    1.00000000 -0.011752230 -0.020279218
+firm -0.01175223  1.000000000 -0.009825009
+foo  -0.02027922 -0.009825009  1.000000000
 > 
 > 
 > 
@@ -664,11 +664,11 @@ foo  -0.02027607 -0.009792208  1.000000000
 > #both bootstrap and computed se:
 > cbind(getfe(est,ef=efactory(est,'ref'),se=TRUE), fse=fixedse(est))
        effect obs comp fe idx         se        fse
-f.1 0.9354485 212    1  f   1 0.06876413 0.07239479
-f.2 1.9504291 208    1  f   2 0.07263999 0.07311460
-f.3 3.0979605 206    1  f   3 0.07552078 0.07351750
-f.4 3.9299698 180    1  f   4 0.07603677 0.07872855
-f.5 5.0332818 194    1  f   5 0.07149293 0.07567327
+f.1 0.9354485 212    1  f   1 0.06752416 0.07239479
+f.2 1.9504291 208    1  f   2 0.06982697 0.07311460
+f.3 3.0979605 206    1  f   3 0.07683635 0.07351750
+f.4 3.9299698 180    1  f   4 0.08075644 0.07872855
+f.5 5.0332818 194    1  f   5 0.07520879 0.07567327
 > #compare with lm:
 > summary(lm(y ~x+f-1))
 
@@ -1060,24 +1060,24 @@ Levels: 1 2 3
 [1] 0.6807816+0i
 >   # mctrace is not really useful for small problems.
 >   mctrace(fun,ncol(A),tol=0.05)
-[1] 0.6560292
+[1] 0.6539472
 attr(,"sd")
-[1] 0.03258954
+[1] 0.03255193
 attr(,"iterations")
-[1] 10868
+[1] 10892
 >   # try a larger problem (3000x3000):
 >   f1 <- factor(sample(1500,3000,replace=TRUE))
 >   f2 <- factor(sample(1500,3000,replace=TRUE))
 >   fl <- list(f1,f2)
 >   mctrace(fl,tol=-5)
-[1] 512.4893
+[1] 512.6789
 attr(,"sd")
-[1] 4.026014
+[1] 4.309791
 attr(,"iterations")
-[1] 44
+[1] 42
 >   # exact:
 >   length(f1) - nlevels(f1) - nlevels(f2) + nlevels(compfactor(fl))
-[1] 507
+[1] 509
 > 
 > 
 > 
@@ -1151,16 +1151,16 @@ x2 -0.12240804  1.224730
 0.2779329 
 > 
 > # Here's a multivalued, vectorized example
-> nlexpect(est, rbind(a=x1*x2 < pt1, b=x1*x2 > 0), vectorized=TRUE)
+> nlexpect(est, rbind(a=x1*x2 < pt1, b=x1*x2 > 0), vectorize=TRUE)
          x1
-a 0.8653844
-b 0.5324968
+a 0.2514457
+b 0.0883645
 > ## No test: 
 > 
 > # Non-linear test:
 > 
 > # A simple one, what's the probability that product x1*x2 is between 0 and |E(x1)|?
-> nlexpect(est, x1*x2 > 0 & x1*x2 < abs(pt1), vectorized=TRUE)
+> nlexpect(est, x1*x2 > 0 & x1*x2 < abs(pt1), vectorize=TRUE)
        x1 
 0.3977857 
 > # Then a more complicated one with the expected value of a polynomal in the coefficients
@@ -1230,14 +1230,14 @@ b 0.5324968
 > # estimate the covariance matrix of the fixed effects
 > fevcov(est, alpha)
               id        firm         foo
-id    0.95289256 -0.04488015  0.03963543
-firm -0.04488015  5.53609420 -0.01160888
-foo   0.03963543 -0.01160888  9.15619224
+id    0.95000141 -0.04658861  0.03880143
+firm -0.04658861  5.53028926 -0.01080117
+foo   0.03880143 -0.01080117  9.15858839
 attr(,"bias")
                id         firm          foo
-id    0.088877778 -0.005940383 -0.004116515
-firm -0.005940383  0.082297201 -0.001586703
-foo  -0.004116515 -0.001586703  0.051505257
+id    0.091768924 -0.004231931 -0.003282517
+firm -0.004231931  0.088102140 -0.002394416
+foo  -0.003282517 -0.002394416  0.049109103
 > # estimate variances of the diagonal
 > varvars(est, alpha)
 [1] 0.01018015 0.05503299 0.09774750
@@ -1289,15 +1289,15 @@ F-statistic:  5219 on 2 and 9997 DF,  p-value: < 2.2e-16
    0.9019639    0.5757362    3.0000000    0.9019611    0.1919121 9997.0000000 
 attr(,"formula")
 ~x - 1 | x2 + 0.2 | `(Intercept)`
-<environment: 0x556e19d899e8>
+<environment: 0x55f2215c27d0>
 > # The Delta-method coincides when the function is linear:
 > waldtest(est, function(x) x - c(0, 1, -0.2))
            p         chi2          df1          p.F            F          df2 
    0.9019639    0.5757362    3.0000000    0.9019611    0.1919121 9997.0000000 
 attr(,"formula")
 function (x) 
 x - c(0, 1, -0.2)
-<bytecode: 0x556e18a54880>
+<bytecode: 0x55f2221a8b78>
 > 
 > 
 > 
@@ -1307,7 +1307,7 @@ x - c(0, 1, -0.2)
 > cleanEx()
 > options(digits = 7L)
 > base::cat("Time elapsed: ", proc.time() - base::get("ptime", pos = 'CheckExEnv'),"\n")
-Time elapsed:  27.778 3.532 26.254 0 0 
+Time elapsed:  28.359 0.921 15.406 0 0 
 > grDevices::dev.off()
 null device 
           1