0
head(jobshop)
  X totalcost units goal.sd weight stamp chisel detail rush labor cost   lost  manager room.temp music shift mach.hrs plant breakdown rework
1 1  90751.53   423     0.1   4.48     4      7     No  Yes  1.47 4.71 0.8317     Alan     74.71  None     2    1.277   Old         1  0.114
2 2 100456.65   554     1.0   4.35     2      3    Yes   No  1.26 4.82 0.4951    Devon     75.37   Pop     1    1.317   New         0  0.000
3 3 128574.01   607     0.5   5.00     2      4     No   No  1.20 5.32 0.5584 Beatrice     75.29  None     1    1.071   Old         0  0.101
4 4  73996.67   347     1.0   5.39     2      4     No   No  1.46 5.44 0.4562  Ebrahim     75.27  Soul     1    1.375   New         0  0.000
5 5  98494.52   510     1.0   4.80     2      4     No   No  1.23 4.98 0.5018  Ebrahim     75.31  Soul     1    1.455   New         0  0.000
6 6  66745.85   419     0.5   3.99     2      4     No   No  1.21 4.60 0.4100  Ebrahim     75.38  Soul     1    1.065   New         0  0.000

my.fit <-lm(totalcost ~ units + goal.sd + weight + stamp + chisel + detail + rush + labor + cost +  lost + manager + room.temp + music + shift + mach.hrs + plant + breakdown + rework, data=jobshop)

goalfact <- as.factor(my.fit$goal.sd) 
stampfact <- as.factor(my.fit$stamp)
chiselfact <- as.factor(my.fit$chisel)
detailfact <- as.factor(my.fit$detail)
rushfact <- as.factor(my.fit$rush)
mgrfact <- as.factor(my.fit$manager)
shiftfact <- as.factor(my.fit$shift)
plntfact <- as.factor(my.fit$plant)
summary(my.fit)

Call:
lm(formula = totalcost ~ units + goal.sd + weight + stamp + chisel + 
    detail + rush + labor + cost + lost + manager + room.temp + 
    music + shift + mach.hrs + plant + breakdown + rework, data = jobshop)

Residuals:
    Min      1Q  Median      3Q     Max 
-118706   -2007     286    2766   18836 

Coefficients: (5 not defined because of singularities)
                  Estimate Std. Error t value Pr(>|t|)    
(Intercept)     -1.065e+05  1.493e+04  -7.129 3.73e-12 ***
units            1.913e+02  2.579e+00  74.151  < 2e-16 ***
goal.sd         -1.519e+03  1.208e+03  -1.257  0.20946    
weight           6.554e+03  5.787e+02  11.326  < 2e-16 ***
stamp           -1.412e+03  7.177e+02  -1.967  0.04972 *  
chisel           1.120e+02  5.057e+02   0.221  0.82487    
detailYes        6.880e+02  1.071e+03   0.643  0.52079    
rushYes          8.511e+02  9.893e+02   0.860  0.39006    
labor            1.112e+04  3.842e+03   2.894  0.00398 ** 
cost             5.920e+03  9.230e+02   6.413 3.40e-10 ***
lost             1.775e+04  6.871e+03   2.583  0.01010 *  
managerBeatrice  1.751e+03  1.235e+03   1.417  0.15714    
managerCarl     -1.137e+04  1.678e+03  -6.774 3.67e-11 ***
managerDevon    -1.164e+04  1.721e+03  -6.767 3.84e-11 ***
managerEbrahim  -1.107e+04  1.679e+03  -6.597 1.11e-10 ***
room.temp        2.108e+01  1.817e+02   0.116  0.90765    
musicPop                NA         NA      NA       NA    
musicRock               NA         NA      NA       NA    
musicSoul               NA         NA      NA       NA    
shift                   NA         NA      NA       NA    
mach.hrs         2.683e+04  1.870e+03  14.353  < 2e-16 ***
plantOld                NA         NA      NA       NA    
breakdown       -3.701e+01  9.261e+02  -0.040  0.96814    
rework          -8.327e+03  1.121e+04  -0.743  0.45795    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 8361 on 481 degrees of freedom
Multiple R-squared:   0.93, Adjusted R-squared:  0.9274 
F-statistic: 355.2 on 18 and 481 DF,  p-value: < 2.2e-16

Hi, I am taking a first course in regression and was trying to fit the above model. I tried to factor or make dummy variables of a few of the predictors but somehow I am getting the Coefficients: (5 not defined because of singularities) and NA for music and breakdown. Please tell me what I'm doing wrong. I am stuck!

horseoftheyear
  • 508
  • 6
  • 12
Sidney Bookeart
  • 1
  • 1
  • 1
  • 1
  • I guess there is some collinearity in your data and based on the variable names which give the `NA`s, possibly some dummy variables are misspecified. What output dus `cor()` give for the variables? Also given the large size of some of your coefficients you might consider rescaling some of the variables. But that just a detail. – horseoftheyear Nov 22 '15 at 17:57

1 Answers1

1
  1. You ran the regression before factoring anything. Your my.fit is still using unfactored data (the goal.sd instead of the goalfact).
  2. You aren't using your factorized variables in your regression. my.fit is still referencing the old data.

Try running this instead: 

    goalfact <- as.factor(my.fit$goal.sd) 
stampfact <- as.factor(my.fit$stamp)
    chiselfact <- as.factor(my.fit$chisel)
detailfact <- as.factor(my.fit$detail)
    rushfact <- as.factor(my.fit$rush)
mgrfact <- as.factor(my.fit$manager)
    shiftfact <- as.factor(my.fit$shift)
plntfact <- as.factor(my.fit$plant)

    my.fit <-lm(totalcost ~ units + goalfact + weight + stampfact + chiselfact + detailfact + rushfact + labor + cost +  lost + managerfact + room.temp + music + shiftfact + mach.hrs + plantfact + breakdown + rework, data=jobshop)`

You have singularities because some of your variables have perfect collinearity; some of your categorical variables occur together 100% of the time. For example, plant = old happens if and only if music = none / rush = yes if and only if shift = 2. You can (and should) remove one of these variables from the regression unless you can find more data where this doesn't happen.

Kontorus
  • 427
  • 2
  • 10