A simple command to grab coefficients, t-stats, p-values, f-stats, etc from a regression and export them as an easy to use spreadsheet.

lmOut

lmOut <- function(res, file="test.csv", ndigit=3, writecsv=T) {
  # If summary has not been run on the model then run summary
  if (length(grep("summary", class(res)))==0) res <- summary(res)
  
  co <- res$coefficients
  nvar <- nrow(co)
  ncol <- ncol(co)
  f <- res$fstatistic
  formatter <- function(x) format(round(x,ndigit),nsmall=ndigit)
  
  # This sets the number of rows before we start recording the coefficients
  nstats <- 4
  
  # G matrix stores data for output
  G <- matrix("", nrow=nvar+nstats, ncol=ncol+1)
    
  G[1,1] <- toString(res$call)
  
  # Save rownames and colnames
  G[(nstats+1):(nvar+nstats),1] <- rownames(co)
  G[nstats, 2:(ncoll+1)] <- colnames(co)
  
  # Save Coefficients
  G[(nstats+1):(nvar+nstats), 2:(ncol+1)] <- formatter(co)
  
  # Save F-stat
  G[1,2] <- paste0("F(",f[2],",",f[3],")")
  G[2,2] <- formatter(f[1])
  
  # Save F-p value
  G[1,3] <- "Prob > P"
  G[2,3] <- formatter(1-pf(f[1],f[2],f[3]))
  
  # Save R2
  G[1,4] <- "R-Squared"
  G[2,4] <- formatter(res$r.squared)
  
  # Save Adj-R2
  G[1,5] <- "Adj-R2"
  G[2,5] <- formatter(res$adj.r.squared)
  
  print(G)
  if (writecsv) write.csv(G, file=file, row.names=F)
}

lmOut(res)

# First let's generate some fake binary response data (from yesterday's post).
  Nobs <- 10^4
  X <- cbind(cons=1, X1=rnorm(Nobs),X2=rnorm(Nobs),X3=rnorm(Nobs),u=rnorm(Nobs))
  B <- c(B0=-.2, B1=-.1,B2=0,B3=-.2,u=5)
  Y <- X%*%B
  SData <- as.data.frame(cbind(Y, X))

# Great, we have generated our data.  
myres <- lm(Y ~ X1 + X2 + X3, data=SData)

lmOut(myres, file="my-results.csv")

Dear Francis,

Thank you very much indeed for your code : )
When I run the code, it showed an error as below:
Error in ncol + 1 : non-numeric argument to binary operator
How to fix it?

Kind regards,
Wei

Thank you for the code! There is a small typo (ncoll instead of ncol) in line 7 that causes the "Error in ncol+1: non-numeric argument to binary operator". Here is the full code:

lmOut <- function(res, file="test.csv", ndigit=3, writecsv=T) {
# If summary has not been run on the model then run summary
if (length(grep("summary", class(res)))==0) res <- summary(res)
co <- res$coefficients
nvar <- nrow(co)
ncol <- ncol(co)
f <- res$fstatistic
formatter <- function(x) format(round(x,ndigit),nsmall=ndigit)
# This sets the number of rows before we start recording the coefficients
nstats <- 4
# G matrix stores data for output
G <- matrix("", nrow=nvar+nstats, ncol=ncol+1)
G[1,1] <- toString(res$call)
# Save rownames and colnames
G[(nstats+1):(nvar+nstats),1] <- rownames(co)
G[nstats, 2:(ncoll+1)] <- colnames(co)
# Save Coefficients
G[(nstats+1):(nvar+nstats), 2:(ncol+1)] <- formatter(co)
# Save F-stat
G[1,2] <- paste0("F(",f[2],",",f[3],")")
G[2,2] <- formatter(f[1])
# Save F-p value
G[1,3] <- "Prob > P"
G[2,3] <- formatter(1-pf(f[1],f[2],f[3]))
# Save R2
G[1,4] <- "R-Squared"
G[2,4] <- formatter(res$r.squared)
# Save Adj-R2
G[1,5] <- "Adj-R2"
G[2,5] <- formatter(res$adj.r.squared)
print(G)
if (writecsv) write.csv(G, file=file, row.names=F)
}

lmOut(res)

# First let's generate some fake binary response data (from yesterday's post).
Nobs <- 10^4
X <- cbind(cons=1, X1=rnorm(Nobs),X2=rnorm(Nobs),X3=rnorm(Nobs),u=rnorm(Nobs))
B <- c(B0=-.2, B1=-.1,B2=0,B3=-.2,u=5)
Y <- X%*%B
SData <- as.data.frame(cbind(Y, X))

# Great, we have generated our data.
myres <- lm(Y ~ X1 + X2 + X3, data=SData)

lmOut(myres, file="my-results.csv")

# Excel summary exporter

Modified from Smart, Francis

Jean P. Gibert, 2014

lmOut <- function(res, file="test.csv", ndigit=3, writecsv=T) {

If summary has not been run on the model then run summary

if (length(grep("summary", class(res)))==0) res <- summary(res)
co <- res$coefficients
nvar <- nrow(co)
ncoll <- ncol(co)
f <- res$fstatistic
formatter <- function(x) format(round(x,ndigit),nsmall=ndigit)

This sets the number of rows before we start recording the coefficients

nstats <- 4

G matrix stores data for output

G <- matrix("", nrow=(nvar+nstats), ncol=(ncoll+1))
G[1,1] <- toString(res$call)

Save rownames and colnames

G[(nstats+1):(nvar+nstats),1] <- rownames(co)
G[nstats, 2:(ncoll+1)] <- colnames(co)

Save Coefficients

G[(nstats+1):(nvar+nstats), 2:(ncoll+1)] <- formatter(co)

Save F-stat

G[1,2] <- paste("F(",f[2],",",f[3],")")
G[2,2] <- formatter(f[1])

Save F-p value

G[1,3] <- "Prob > P"
G[2,3] <- formatter(1-pf(f[1],f[2],f[3]))

Save R2

G[1,4] <- "R-Squared"
G[2,4] <- formatter(res$r.squared)

Save Adj-R2

G[1,5] <- "Adj-R2"
G[2,5] <- formatter(res$adj.r.squared)
print(G)
write.csv(G, file=file, row.names=F)
}

Thanks for the corrections!

Hi,

This function is great and I appreciate you putting it together. Question that I can't figure out unfortunately (I am a beginner.)

Trying to loop the output for multiple regressions that I'm doing across monthly samples.

This works fine: lmOut(m_2014-01-31,file="my-results.csv")

But the loop below (where monthschar[1]= 2014-01-31) does not work and returns:
Error in res$coefficients : $ operator is invalid for atomic vectors

for (i in 1:length(monthschar)) {
lmOut(paste("m_",monthschar[i],sep="",paste("",sep="")),
file=paste("output",monthschar[i],sep=""),paste(".csv",sep=""))
}

Any idea why this is happening? The silly quotation marks (`) that R put around the models I generated with another for loop are the reason I attached the ugly paste notation.

Thank you!

Hi

Thanks for the function. I encounter this error:
Error in matrix("", nrow = (nvar + nstats), ncol = (ncoll + 1)) :
invalid 'nrow' value (too large or NA)

I am using a linear mixed effect model.
Thanks

thanks for the code ...I am also a beginner and it is very useful... thanks for putting putting some sense into to responder2 on R bloogers... http://www.r-bloggers.com/export-r-results-tables-to-excel-please-dont-kick-me-out-of-your-club/

Just curious, why haven't you made this into a package?

much more canonical than

if (length(grep("summary", class(res)))==0)

would be

if (inherits(res, 'summary.lm'))

Thank you this is very helpful. I did notice that line 7 and line 17 and 19 all need to be consistent "ncol" or "ncoll" .

@MichaelChirico: I change
"if (length(grep("summary", class(res)))==0) res <- summary(res) "
to
" if (inherits(res, 'summary.lm')) res <- summary(res)"

but then the function throws an error:

"Error in matrix("", nrow = nvar + nstats, ncol = ncol + 1) :
invalid 'nrow' value (too large or NA)"

The original version does produce the "my-results.csv" output file.

Hello, many thanks for this :) I'm trying to run it but I'm getting the following error:

Error in grep("summary", class(res)) : object 'res' not found

Apologies if this is something basic, I am very, very new to R!
Thanks

Hi, really helpful! is there any version of this function that works with logistic regressions?

To make it work with plm models, modify code this way:

• change f[1] to f[2]$statistic
• change f[2] to f[3]$parameter["df1"]
• change f[3] to f[3]$parameter["df2"]
• change res$r.squared to res$r.squared["rsq"]
• change res$adj.r.squared to res$r.squared["adjrsq"]

Hi, thanks for writing this helpful function. I'm curious that is there any way to use it when I do the regression under glm function?