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")

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?