fix crash in demeaning when nobs = 1, fixes: #283

NEWS.md

@@ -22,6 +22,8 @@
 - fix bug in the `collinearity` function. Reported by @grlju, #412
 
 - fix bug which allowed the estimation of models with variables from the environment while it shouldn't
+
+- fix bug in the demeaning code when the data set in input is of length 1. Reported by @da-zar with help of @etiennebacher, #283
 
 ## Improvements
 

R/estimation.R

@@ -1594,6 +1594,9 @@ feols = function(fml, data, vcov, weights, offset, subset, split, fsplit, split.
   #
 
   onlyFixef = length(X) == 1 || ncol(X) == 0
+  if(length(y) == 1){
+    onlyFixef = is.null(dim(X)) || ncol(X) == 0
+  }
 
   if(fromGLM){
     res = list(coefficients = NA)
@@ -1671,7 +1674,7 @@ feols = function(fml, data, vcov, weights, offset, subset, split, fsplit, split.
 
         res$convStatus = FALSE
 
-        res$message = paste0("tol: ", signif_plus(fixef.tol), ", iter: ", max(res$iterations))
+        res$message = paste0("tol: ", fsignif(fixef.tol), ", iter: ", max(res$iterations))
 
         if(fromGLM){
           res$warn_varying_slope = msg
@@ -1719,7 +1722,7 @@ feols = function(fml, data, vcov, weights, offset, subset, split, fsplit, split.
   if(mem.clean){
     gc()
   }
-
+  
   if(!onlyFixef){
 
     est = ols_fit(y_demean, X_demean, weights, correct_0w, collin.tol, nthreads,
@@ -1739,13 +1742,9 @@ feols = function(fml, data, vcov, weights, offset, subset, split, fsplit, split.
       all_vars = colnames(X)
 
       if(isFixef){
-        msg = paste0(ifsingle(all_vars, "The only variable ", "All variables"), 
-                     enumerate_items(all_vars, "quote.is", nmax = 3), 
-                     " collinear with the fixed effects. In such circumstances, the estimation is void.")
+        msg = sma("{$(The only variable;All variables)}, {enum.q.3 ? all_vars}, {$are} collinear with the fixed effects. In such circumstances, the estimation is void.")
       } else {
-        msg =  paste0(ifsingle(all_vars, "The only variable ", "All variables"), 
-                      enumerate_items(all_vars, "quote.is", nmax = 3), 
-                      " virtually constant and equal to 0. In such circumstances, the estimation is void.")
+        msg = sma("{$(The only variable;All variables)}, {enum.q.3 ? all_vars}, {$are} virtually constant and equal to 0. In such circumstances, the estimation is void.")
       }
 
       if(IN_MULTI || !warn){
@@ -1761,7 +1760,7 @@ feols = function(fml, data, vcov, weights, offset, subset, split, fsplit, split.
         return(fixest_NA_results(env))
 
       } else {
-        stop_up(msg, up = fromGLM)
+        stop_up(msg, up = 1 * fromGLM)
       }
 
 

src/demeaning.cpp

@@ -156,7 +156,7 @@ class sMat{
   sMat() = delete;
 
 public:
-  sMat(SEXP);
+  sMat(SEXP, bool);
 
   int nrow(){return n;};
   int ncol(){return K;};
@@ -165,7 +165,7 @@ class sMat{
   double operator()(int, int);
 };
 
-sMat::sMat(SEXP x){
+sMat::sMat(SEXP x, bool single_obs = false){
 
   if(TYPEOF(x) == VECSXP){
     // x can be a list of either vectors or matrices
@@ -231,7 +231,7 @@ sMat::sMat(SEXP x){
       K = pdim[1];
     }
 
-    if(n == 1 && K == 1){
+    if(!single_obs && (n == 1 && K == 1)){
       // => absence of data
       n = 0;
       K = 0;
@@ -1797,8 +1797,8 @@ void stayIdleCheckingInterrupt(bool *stopnow, vector<int> &jobdone, int n_vars,
 // Loop over demean_single
 // [[Rcpp::export]]
 List cpp_demean(SEXP y, SEXP X_raw, SEXP r_weights, int iterMax, double diffMax, SEXP r_nb_id_Q,
-           SEXP fe_id_list, SEXP table_id_I, SEXP slope_flag_Q, SEXP slope_vars_list,
-           SEXP r_init, int nthreads, bool save_fixef = false){
+                SEXP fe_id_list, SEXP table_id_I, SEXP slope_flag_Q, SEXP slope_vars_list,
+                SEXP r_init, int nthreads, bool save_fixef = false){
   // main fun that calls demean_single
   // preformats all the information needed on the fixed-effects
   // y: the dependent variable
@@ -1827,6 +1827,20 @@ List cpp_demean(SEXP y, SEXP X_raw, SEXP r_weights, int iterMax, double diffMax,
     n_obs = m_X.nrow();
   }
   bool useX = n_vars_X > 0;
+
+  if(n_obs == 0 || n_obs == 1){
+    // The data set is of length 1!!!!
+    n_obs = 1;
+
+    m_y = sMat(y, true);
+    n_vars_y = m_y.ncol();
+    useY = true;
+
+    m_X = sMat(X_raw, true);
+    n_vars_X = m_X.ncol();
+    useX = n_vars_X > 0;
+
+  }
 
   int n_vars = n_vars_y + n_vars_X;
 
@@ -1836,7 +1850,8 @@ List cpp_demean(SEXP y, SEXP X_raw, SEXP r_weights, int iterMax, double diffMax,
   bool saveInit = ((isInit || init[0] != 0) && Q > 1) || init[0] == 666;
 
   // Creating the object containing all information on the FEs
-  FEClass FE_info(n_obs, Q, r_weights, fe_id_list, r_nb_id_Q, table_id_I, slope_flag_Q, slope_vars_list);
+  FEClass FE_info(n_obs, Q, r_weights, fe_id_list, r_nb_id_Q, table_id_I, slope_flag_Q, 
+                  slope_vars_list);
   int nb_coef_T = FE_info.nb_coef_T;
 
   // output vector: (Note that if the means are provided, we use that vector and will modify it in place)
@@ -1971,7 +1986,6 @@ List cpp_demean(SEXP y, SEXP X_raw, SEXP r_weights, int iterMax, double diffMax,
   res["X_demean"] = X_demean;
 
 
-
   if(is_y_list && useY){
     List y_demean(n_vars_y);
 

tests/fixest_tests.R

@@ -256,6 +256,11 @@ test(feols(y ~ x1 | factor(fe1^fe2), base), "err")
 
 res = feols(y ~ x1 | bin(x2, "bin::1")^fe1 + fe1^fe2, base)
 
+# 1 obs (after FE removal) estimation
+base = data.frame(y = c(1, 0), fe = c(1, 2), x = c(1, 0))
+test(fepois(y ~ x | fe, base), "err")
+# no error
+res = fepois(y ~ 1 | fe, base)