Skip to content
A function for rounding values in R markdown
R
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
README.md
cutoff.R

README.md

A function for rounding values in R Markdown

In R Markdown one is often faced with situations where a value that can vary substantially is automatically pulled out. For example, we might want to pull a p-value out of an linear model object. It's a chore to have to manually edit p-values so that they conform with reporting conventions (e.g. p < .001).

At the end of this document is a function cutoff that deals with these sort of cases by taking numeric value and returning "> [upper]", "< [lower]" or "= [exact value to 2 sf]" depending on the value. Thus for a p-value of, say, 6.447569e-05 the function would return "< .001". Below are some examples of what the standard function would return with different inputs.

Input Output
0.674 "= .67"
0.027 "= .027"
0.00053 "< .001"

By default the upper cut-off is set to 1 meaning it isn't used for a p-value and the lower cut-off is set to 0.001. However, these cut-offs can be adjusted with upper and lower arguments, respectively. The sig argument sets the number of significant figures (default = 2) to report for exact values between the upper and lower cut-offs. The return.num argument is false by default. If it's set to true then the "=" or "<" aren't included in the returned string. E.g. ".001" rather than "< .001" or .032 rather than "= .032". The lead.zero argument defaults to FALSE and returns values with the leading zero, e.g., .004 rather than 0.004. This is the recommended formatting for numbers than cannot take values larger than 1, such a p-values. The arguments of the function are summarised below.

Argument Info
x a vector
upper value above which "> [upper]" is returned
lower value below which "< [lower]" is returned
sig number of significant figures to round values between upper and lower to
return.num logical, whether to return just the number (i.e. not include the = or >)
lead.zero logical, include leading zero (e.g. 0.01)
equal.sign logical, include the equal sign for values between upper and lower

The equal.sign argue might be useful when using the function in a table. This would give you ".16" instead of "= .16" for, say, a p value in a table but the "< .001"s would stay the same.

The file cutoff.R contains different version of the function with defaults for different statistics. These could, of course, be replicated by changing the upper and lower arguments. However, it might be easier and more readable to have a different versions of the function for each type of statistic. The default arguments of the different version are summarised below. cutoff2 uses round() rather than signif() so that it's suited to large values with decimals, such as percentages. It has a dp rather than a sig arguments which defaults to 1 decimal place, e.g. 56.9. cutoff_BF has both dp and sig arguments for values above and below 1 respectively. The defaults allow you to round values greater than 1 to one decimal place while values less than 1 have two significant figures.

Function name upper lower sig / dp return.num lead.zero equal.sign Notes
cutoff 1 0.001 2 FALSE FALSE TRUE
cutoff_Rsq 1 0.01 2 FALSE FALSE TRUE
cutoff_beta 1 0.0001 2 FALSE TRUE TRUE
cutoff2 100 0.01 2 FALSE TRUE TRUE Uses round() not signif()
cutoff_BF 1000 0.001 2 / 1 FALSE TRUE TRUE dp arg rather than sig

All the function are vectorised so can be applied to a column of values.

If this function looks a bit loop heavy for R then my only defence is I was doing a lot of Python at the time.

#==============================================================================
# Standard version: uses signif(), defaults for p-values
#==============================================================================

cutoff = function(x, upper = 1, lower = 0.001, sig = 2, return.num = FALSE, 
                  lead.zero = FALSE, equal.sign = TRUE) {
  y = NULL
  z = NULL
  
  if (lead.zero == FALSE) {
    s = 2
  } else {
    s = 1
  }
  
  for(i in seq_along(x)) {
    if(abs(x[i]) < lower) {
      y[i] = paste("<", substring(as.character(lower), s), sep = " ") # substring of a character used to get .001 rather than 0.001
      z[i] = substring(as.character(lower), s)
      
    } else if(abs(x[i]) > upper & upper < 1) {
      y[i] = paste(">", substring(as.character(upper), s), sep = " ")
      z[i] = substring(as.character(upper), s)
      
    } else if(abs(x[i]) > upper & upper >= 1) {
      y[i] = paste(">", substring(as.character(upper), 1), sep = " ")
      z[i] = substring(as.character(upper), 1)
      
    } else {
      y[i] = paste("=", substring(as.character(signif(x[i], sig)), s), sep = " ")
      z[i] = substring(as.character(signif(x[i], sig)), s)
    }
  }
  
  if (return.num == FALSE & equal.sign == TRUE) {
    return(y)
    
  } else if(equal.sign == FALSE) {
    for(j in seq_along(y)) {
      if(substring(y[j], 1, 1) == "=") {
        y[j] <- substring(y[j],3)
      }
    }
    return(y)
    
  } else if(return.num == TRUE) {
    return(z)
  }
}
You can’t perform that action at this time.