Recommended R functions: Difference between revisions

Revision as of 16:09, 30 December 2011

This page is a knowledge crystal of subtype method. The page identifier is Op_en5339
Moderator:Jouni (see all)
This page is a stub. You may improve it into a full page.
Upload data {{#opasnet_base_link:Op_en5339}}

Recommended R functions describes good practices for writing R code. The code should be short, straightforward to understand, efficient to run, and similar to everyone else's code.

Question

What are good practices for writing R code? The code should be short, straightforward to understand, efficient to run, and similar to everyone else's code.

Answer

Recommended structures

When possible, use data.frames rather than arrays, tables or lists.
Standard columns for data.frames:
- id: row identifier in the res table of Opasnet Base. Usually not needed and this can be sliced away.
- obs: row identifier in a data table or one uploaded piece of data. Technically, a piece of data that has the same series_id in Opasnet Base.
- iter: identifier of the iteration in a Monte Carlo simulation.
- There are several standard indices in Opasnet Base such as Year, Sex, Age, Lo (longitude), La (latitude),... If possible, use these names. For a full reference, see Opasnet Base Indices.
- Result: the actual result of the object, typically numeric. There are also other columns used for results, namely
  - Result.Text: in Opasnet Base, results that are in text format are stored in this column, and
  - Freq: when tapply() function is used, the summarised result is given in Freq column.
- Unit: If the unit can be different at different rows, a separate column Unit is needed.
- Description: Description can contain any descriptive information about the row. It is not used in calculations.
For health impacts, there are standard tables to be used. For details, see Health impact assessment.
For time tracking of working hours, there are standard tables to be used. For details, see op_fi:Aikakone.
For listing several similar tables that should be bound rowwise, use standard tables described in Using summary tables.

Recommended generic functions

**Recommended functions and operations.**
What to do	Functions to use routinely	Functions to avoid except in special cases	Examples and description
Manipulate data	data.frame	array
Draw raphs	ggplot, plot		ggplot requires library(ggplot2)
Summarise data along a criterion	tapply
Join two data.frames	merge	IntArray
Add rows to a data.frame	rbind
Add columns to a data.frame	cbind
Transform a table from long to wide or vice versa	reshape
Get data from Opasnet Base	op_baseGetData		Requires library(OpasnetBaseUtils) > op_baseGetData("opasnet_base", "Op_en4523")[, -c(1,2)] # Gets the object with identifier Op_en4523 and slices columns 1 and 2 (id, obs) away.
Get index values from Opasnet Base	op_baseGetLocs		Requires library(OpasnetBaseUtils)
Write data to Opasnet Base	op_baseWrite		Requires library(OpasnetBaseUtils). Only works from a THL computer, not R-tools
Slicing R objects	data[rows, cols], data$col		This is not a function but rather a list of practical ways of slicing an object.
	match
	ifelse
	is.na
	colnames
Convert between data types	as.numeric, as.character, as.factor
Convert between object types	as.data.frame, as.table

Recommended tailored functions

Code tailoredfunctions:

multivarplot produces a graph for multiple variables along the same X axis (usually a timeline).

+ Show code - Hide code

##### multivarplot produces a graph for multiple variables along the same X axis (usually a timeline). Parameters:
#     a: Data.frame that has three columns: DateTime = x axis value, TagName = name of the variable, and Value = y axis value for the variable.
#     precision: a smoothing parameter (0 = no smoothing)
#     timeline: TRUE if DateTime has POSIXct format, FALSE if real number.

multivarplot <- function(a, precision = 0, timeline = FALSE) {
	par(mar=c(5, length(levels(a$TagName)) * 3.5 + 1.5, 4, 4) + 0.1)
	
	x <- 0
	
	for (i in levels(a$TagName)) {
		if(i != levels(a$TagName)[1]) par(new = TRUE)
		plot(if(precision == 0) list(x = a$DateTime[a$TagName == i], y = a$Value[a$TagName == i]) else loess.smooth(a$DateTime[a$TagName == i], a$Value[a$TagName == i], 
			degree = 1, span = precision), axes = FALSE, xlab = "", ylab = "", type = "l", col = rainbow(length(levels(a$TagName)))[x + 1], main = "", 
			xlim = c(min(a$DateTime), max(a$DateTime)), ylim = c(min(a$Value[a$TagName == i]) - sd(a$Value[a$TagName == i]) * 
			0.1, max(a$Value[a$TagName == i]) + sd(a$Value[a$TagName == i]) * 0.1))
		axis(2, col = rainbow(length(levels(a$TagName)))[x + 1],lwd = 2, line = x * 3.5)
		mtext(2, text = i, line = x * 3.5 + 2, col = rainbow(length(levels(a$TagName)))[x + 1])
		x <- x + 1
	}
	
	if(timeline){axis.POSIXct(1, a$DateTime)} else {axis(1, a$DateTime)}
	mtext("Time", side = 1, col = "black", line = 2)
}

Rationale

Based on experience and testing.

References

Related files