R help - Aug 2010 - "2 not defined because of singularities" appearing after introducing Fixed Effects

The set runs fine without the fixed effects. However, once I add the
"+factor(HostCode)"-part, it throws out two variables. The
africa-dummy
thrown out certainly does not exhibit perfect multicollinearity, I checked
that. The litrate variable is continuous and therefore cannot be perfectly
related to the FE-dummies introduced by R.

This is the used code:


> cdmoutcome<- lm(log(value)~#factor(year)
>               +factor(HostCode)
>               +log(gdppcpppconst)+log(gdppcpppconstAII)
>               +log(co2eemisspc)+log(co2eemisspcAII)
>               +fdiboth
>               +odapartnertohost
>               +infraindex
>               +litrate
>               +africa
>               +imr
>                  , data=cdmdata2, subset=zero==1)
> summary(cdmoutcome
> 
These are the results (please expand to see the relevant part):


> Coefficients: (2 not defined because of singularities)
>                         Estimate Std. Error t value Pr(>|t|)    
> (Intercept)            1.144e+01  3.077e+01   0.372   0.7103    
> factor(HostCode)7     -3.296e+00  4.617e+00  -0.714   0.4759    
> factor(HostCode)19    -1.242e+01  8.764e+00  -1.417   0.1575    
> factor(HostCode)20     1.548e+00  4.087e+00   0.379   0.7052    
> factor(HostCode)23    -4.183e+00  3.786e+00  -1.105   0.2702    
> factor(HostCode)29    -1.473e+01  1.168e+01  -1.261   0.2083    
> factor(HostCode)35     7.631e-01  7.870e-01   0.970   0.3330    
> factor(HostCode)36     7.700e+00  4.340e+00   1.774   0.0771 .  
> factor(HostCode)37    -6.549e+00  4.844e+00  -1.352   0.1774    
> factor(HostCode)42    -1.020e+01  5.074e+00  -2.010   0.0453 *  
> factor(HostCode)51     1.738e+00  3.206e+00   0.542   0.5882    
> factor(HostCode)52     1.032e+00  2.742e+00   0.376   0.7069    
> factor(HostCode)53     3.210e+00  2.768e+00   1.160   0.2471    
> factor(HostCode)54    -8.851e+00  6.340e+00  -1.396   0.1637    
> factor(HostCode)60     2.146e+00  3.531e+00   0.608   0.5438    
> factor(HostCode)65    -7.436e+00  5.958e+00  -1.248   0.2129    
> factor(HostCode)70    -9.056e+00  6.711e+00  -1.349   0.1782    
> factor(HostCode)75    -5.239e+00  5.696e+00  -0.920   0.3585    
> factor(HostCode)78    -4.482e+00  5.185e+00  -0.864   0.3881    
> factor(HostCode)79     4.259e-01  3.777e+00   0.113   0.9103    
> factor(HostCode)84     2.035e+00  3.374e+00   0.603   0.5468    
> factor(HostCode)94    -2.008e+01  1.315e+01  -1.528   0.1276    
> factor(HostCode)106    6.563e+00  3.027e+00   2.168   0.0309 *  
> factor(HostCode)113    5.919e+00  1.010e+00   5.858 1.23e-08 ***
> factor(HostCode)115   -2.429e+00  4.274e+00  -0.568   0.5702    
> factor(HostCode)117    1.240e+01  5.779e+00   2.146   0.0327 *  
> factor(HostCode)119   -8.822e-01  4.118e+00  -0.214   0.8305    
> factor(HostCode)123   -2.530e+01  1.616e+01  -1.566   0.1185    
> factor(HostCode)126   -5.210e+00  6.782e+00  -0.768   0.4430    
> factor(HostCode)128   -7.044e+00  8.246e+00  -0.854   0.3937    
> factor(HostCode)132   -3.570e+00  6.508e+00  -0.548   0.5838    
> factor(HostCode)134   -1.756e+00  3.905e+00  -0.450   0.6534    
> factor(HostCode)135   -3.403e+00  7.444e+00  -0.457   0.6479    
> factor(HostCode)137   -7.738e+00  6.175e+00  -1.253   0.2111    
> factor(HostCode)138   -1.066e+01  7.116e+00  -1.498   0.1351    
> factor(HostCode)160    1.251e+00  4.889e+00   0.256   0.7982    
> factor(HostCode)162   -1.337e+01  8.742e+00  -1.529   0.1273    
> factor(HostCode)171   -1.804e+00  2.404e+00  -0.750   0.4537    
> factor(HostCode)176    3.873e+00  2.711e+00   1.428   0.1542    
> factor(HostCode)180   -3.186e+01  1.734e+01  -1.837   0.0671 .  
> factor(HostCode)185   -2.367e+01  5.831e+00  -4.059 6.30e-05 ***
> factor(HostCode)189   -4.209e+00  5.328e+00  -0.790   0.4302    
> log(gdppcpppconst)     1.116e+00  3.849e+00   0.290   0.7721    
> log(gdppcpppconstAII) -7.209e-01  7.850e-01  -0.918   0.3592    
> log(co2eemisspc)      -1.149e+01  6.230e+00  -1.845   0.0661 .  
> log(co2eemisspcAII)   -2.182e-01  3.851e-01  -0.567   0.5714    
> fdiboth                1.707e-04  1.055e-04   1.619   0.1066    
> odapartnertohost       2.886e-03  1.354e-03   2.132   0.0338 *  
> infraindex             4.394e+01  9.097e+00   4.830 2.18e-06 ***
> litrate                       NA         NA      NA       NA    
> africa                        NA         NA      NA       NA    
> imr                    1.470e+00  7.173e-01   2.049   0.0414 *
> 
Does anyone know how to solve this problem so I can use the FE and still
keep the variables in? (Or what I made have wrong on a technical level?)
-- 
View this message in context:
http://r.789695.n4.nabble.com/2-not-defined-because-of-singularities-appearing-after-introducing-Fixed-Effects-tp2327047p2327047.html
Sent from the R help mailing list archive at Nabble.com.

Sorry, please nevermind. It seems to have been an econometrical problem after
all.

(A variable consisting of 2 or more of the 150 Country-Dummies for the fixed
effects causes perfect multicollinearity as well. So does a variable that
differs over cross-sections but not over periods, seemingly. I was not aware
before...)
-- 
View this message in context:
http://r.789695.n4.nabble.com/2-not-defined-because-of-singularities-appearing-after-introducing-Fixed-Effects-tp2327047p2328686.html
Sent from the R help mailing list archive at Nabble.com.