[R] function that join my model & these coefficients

[Jim Lemon]

On Sun, 27 Jul 2014 02:08:25 AM Maede Nouri wrote:
> hello 
> I am new to R language. I fitted a linear model and my output has 
about 300
> coefficients. I need to definition a function that join my model & 
these
> coefficients ! this is difficult because number of coefficients is many.  
I
> think there is not a provided query for this. I also used "fitted(fit) #
> predicted values" but it can't help me to recieve my goal. please 
help
> me in finding the functions and merge my model & its coefficients.
> 
> 
> data2<-subset(data1,pd2<660000)
> 
> > sapply(data2,mode)
> 
>         aid         act        acid      id_new      userid         pd1    
>     pd2         pd3         pd4         pd5       freq1       freq2      
> freq3  "numeric"   "numeric"   "numeric"   "numeric" "character"
> "character" "character" "character" "character"   "numeric"   
"numeric"  
> "numeric"   "numeric" freq4       freq5
>   "numeric"   "numeric" 
> 
> > fit <- lm(act
> > 
~freq1+freq2+freq3+freq4+freq5+pd1*freq1+pd2*freq2+pd3*freq3+pd4*freq4+pd
> > 5*freq5-1,data=data2) summary(fit)
> 
> Call:
> lm(formula = act ~ freq1 + freq2 + freq3 + freq4 + freq5 + pd1 * 
>     freq1 + pd2 * freq2 + pd3 * freq3 + pd4 * freq4 + pd5 * freq5 - 
>     1, data = data2)
> 
> Residuals:
>     Min      1Q  Median      3Q     Max 
> -26.905  -2.843   0.000   1.606  33.059 
> 
> Coefficients: (233 not defined because of singularities)
>                   Estimate Std. Error t value Pr(>|t|)    
> freq1            1.293e-01  1.753e-01   0.738 0.461206    
> freq2            2.016e-01  3.310e-01   0.609 0.542809    
> freq3           -4.816e+00  2.220e+00  -2.169 0.030795 *  
> freq4            2.395e-01  1.751e+00   0.137 0.891272    
> freq5           -9.289e+00  6.110e+00  -1.520 0.129394    
> pd1630000        5.625e+00  5.978e+00   0.941 0.347445    
> pd1646000        7.082e+00  1.410e+01   0.502 0.615714    
> pd1648000        1.275e+00  4.240e+01   0.030 0.976027    
> pd1651000        3.404e+00  5.694e+00   0.598 0.550352    
> pd1656000       -8.177e+00  1.017e+01  -0.804 0.421906    
> pd1665000        3.795e+00  5.649e+00   0.672 0.502231    
> pd1666000        9.805e+00  5.857e+00   1.674 0.095058 .  
> pd2651000        4.790e+00  6.122e+00   0.782 0.434510    
> pd2656000        1.754e+00  5.620e+00   0.312 0.755221    
> pd2659000        4.187e-01  9.814e+00   0.043 0.965996    
> pd3630000       -7.989e+00  5.629e+00  -1.419 0.156780    
> pd3646000       -1.638e+00  1.772e+01  -0.092 0.926444    
> pd3648000       -4.021e+00  7.725e+00  -0.521 0.602998    
> pd3651000       -4.848e+00  6.400e+00  -0.758 0.449243    
> pd3656000       -6.105e-01  8.732e+00  -0.070 0.944303    
> pd3659000        4.474e+00  7.483e+00   0.598 0.550296    
> pd3663000        1.299e+02  4.969e+01   2.614 0.009360 ** 
> pd3737082               NA         NA      NA       NA    
> pd3737110        9.847e+01  3.120e+01   3.156 0.001744 ** 
> pd3738240        5.675e+00  1.223e+01   0.464 0.643073    
> pd4646000        8.603e+00  1.751e+01   0.491 0.623605    
> pd4648000        3.040e+00  4.848e+00   0.627 0.531064    
> pd4651000        9.141e-01  5.243e+00   0.174 0.861689    
> pd5              1.132e-06  3.440e-06   0.329 0.742355    
> freq1:pd1646000         NA         NA      NA       NA    
> freq1:pd1648000 -1.786e+00  1.267e+01  -0.141 0.888002    
> freq2:pd2646000  1.369e+01  1.666e+01   0.822 0.411865    
> freq2:pd2648000  1.396e+00  2.172e+00   0.643 0.520933    
> freq3:pd3693000  3.270e+00  2.557e+00   1.279 0.201910    
> freq3:pd3694000  7.476e+00  3.571e+00   2.094 0.037035 *  
> freq3:pd3699000  5.310e+00  2.187e+00   2.429 0.015679 *  
> 
> 
Hi Maede,
The number of singularities strongly suggests that you have way too 
few data points to attempt a model with that many interaction terms. 
Even so, such a model would be exceedingly difficult to interpret. The 
numbered variables wouldn't be repeated observations, would they?

Jim