R loop through the independent variables in lm function

Question

I am having a problem with building lm function based on many independent variables in for loop. 14 different independent variables (x1, x2, x3 ..., x14) are created in each for loop and as a result the name of the variables (strings) are saved in vector 'independent_variables'. For dependent variable y1, I would like to build the lm function lm(y1 ~ x1 + x2 + x3 + x4 + x5 + x6 + x7 + x8 + x9 + x10 + x11 + x12 + x13 + x14)

I've tried to paste the elements of the list and typed it in lm function, but it doesn't seem to recognize this as a formula.

 for (j in 1:length(num)) { 
      nam <- paste("x",j, sep="")
      assign(nam, vec)
      independent_variables <- c(independent_variables, nam)
    }    

> independent_variables
  [1] "x1"  "x2"  "x3"  "x4"  "x5"  "x6"  "x7"  "x8"  "x9"  "x10" "x11" 
      "x12" "x13" "x14"

they are independent variables of the linear regression function and each element a matrix which has 318 rows in 1 column. Also, for the dependent variable y1, I have a matrix which has the same dimension.

> x1
                                                    COAD_65
ACCx_025FE5F8_885E_433D_9018_7AE322A92285_X034_S09 -0.368827920
ACCx_2A5AE757_20D5_49B6_95FF_CAE08E8197A0_X012_S05 -0.418133754
ACCx_3D0CD3BD_3960_46FB_92C3_777F11CCD0FC_X011_S06 -0.885246719
ACCx_4D0D43F5_D8F0_4735_92D5_F40E321C7A05_X010_S09 -0.908954868
ACCx_81A262BD_3078_4BDB_8EB1_30DD6D7948C3_X027_S03 -0.284544506
ACCx_B6E6F014_A599_4A58_A7A5_1F748471D662_X013_S12 -0.991800815
ACCx_B901534B_5E93_475A_91E7_B2DB7DFE98A5_X020_S02 -0.538162178
ACCx_EDEB779F_A603_479D_AAFE_428BC7E4B8DB_X038_S03 -0.462774125
...

UCEC_BDFE8123_081E_49AF_930B_2371D8DEC261_X030_S01 -1.032249118
UCEC_C335297F_2D63_4973_9182_FA18C28E001E_X037_S04 -0.550676273
UCEC_D820B024_6B3B_4B5B_866E_F9A8139C270B_X039_S09 -0.036913872


> y1
 TCGA-OR-A5K8-01A TCGA-PK-A5H8-01A TCGA-OR-A5J3-01A TCGA-OR-A5J6-01A TCGA- 
 OR-A5KX-01A TCGA-OR-A5J2-01A 
 0.000000000      0.000000000      0.013752216      0.000000000      
 0.000000000      0.000000000 
 TCGA-OR-A5J9-01A TCGA-OR-A5JZ-01A TCGA-PA-A5YG-01A TCGA-CU-A3YL-01A TCGA- 
 GD-A3OQ-01A TCGA-CF-A3MI-01A 
 0.009707204      0.000000000      0.000000000      0.000000000      
 0.000000000      0.119174367 
 ...
 TCGA-BL-A13J-01A TCGA-GV-A3JW-01A TCGA-DK-A1AD-01A TCGA-FD-A3SR-01A TCGA- 
 CF-A1HR-01A TCGA-BL-A3JM-01A 
 0.019066953      0.355925504      0.019473742      0.062201816      
 0.081559894      0.243386421

After creating correct lm function, the result should look like this for example

> Call: lm(formula = y1 ~ x1 + x2 + x3 + x4 + x5 + x6 + x7 + x8 + x9 + 
>     x10 + x11 + x12 + x13 + x14)
> 
> Residuals:
>     Min      1Q  Median      3Q     Max 
> -0.6282 -0.1130 -0.0257  0.0491  6.0798 
> 
> Coefficients:
>              Estimate Std. Error t value Pr(>|t|)     (Intercept)  0.054546   0.040219   1.356   0.1759     x1           0.145644   0.035340   4.121 4.66e-05 *** x2           0.005909   0.038020   0.155   0.8766     x3          -0.085892   0.051854  -1.656   0.0985 .   x4           0.032686   0.029443   1.110   0.2677     x5          -0.047268   0.033388  -1.416   0.1577     x6           0.026735   0.032327   0.827   0.4088     x7           0.035673   0.051047   0.699   0.4851     x8           0.037374   0.060258   0.620   0.5355     x9           0.024493   0.053045   0.462   0.6445     x10          0.006623   0.059025   0.112   0.9107     x11         -0.017017   0.034501  -0.493   0.6221     x12          0.032184   0.046235   0.696   0.4868     x13          0.009988   0.033298   0.300   0.7644     x14         -0.017836   0.024505  -0.728   0.4672    
> --- Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
> 
> Residual standard error: 0.3936 on 366 degrees of freedom Multiple
> R-squared:  0.2768,   Adjusted R-squared:  0.2492  F-statistic: 10.01 on
> 14 and 366 DF,  p-value: < 2.2e-16

A few more good ideas here - https://stackoverflow.com/questions/5251507/how-to-succinctly-write-a-formula-with-many-variables-from-a-data-frame . Especially the reformulate function (ex `reformulate(names(anscombe), "y1")`) — M.Viking, Jun 14 '19 at 14:13

G. Grothendieck · Accepted Answer · 2019-06-14T12:47:05.270

1

Using the builtin anscombe data frame:

names(anscombe)
## [1] "x1" "x2" "x3" "x4" "y1" "y2" "y3" "y4"
indep_names <- grep("x", names(anscombe), value = TRUE)

lm(anscombe[c("y1", indep_names)])

or

lm(y1 ~., anscombe[c("y1", indep_names)])

or

fo <- sprintf("y1 ~ %s", paste(indep_names, collapse = "+"))
do.call("lm", c(fo, quote(anscombe)))

or

fo <- reformulate(indep_names, response = "y1")
do.call("lm", c(fo, quote(anscombe)))

In the last two cases we could write lm(fo, anscombe) instead if we don't care that the formula then shows up as just fo in the output.

edited Jun 14 '19 at 12:47

answered Jun 14 '19 at 12:40

G. Grothendieck

254,981
17
203
341

I've tried third one and it worked. Thanks! Seems like lm(fo) would make the same result as it should be, is 'lm(fo)' the right method as well? – na.0o0 Jun 14 '19 at 12:51
lm(fo) will work if the variables are not in a data.frame; however, note that the output would show the formula as `fo` rather than showing the content of `fo` (as mentioned already in the answer). – G. Grothendieck Jun 14 '19 at 12:57

R loop through the independent variables in lm function

1 Answers1