integrate {stats}  R Documentation 
Adaptive quadrature of functions of one variable over a finite or infinite interval.
integrate(f, lower, upper, ..., subdivisions = 100L,
rel.tol = .Machine$double.eps^0.25, abs.tol = rel.tol,
stop.on.error = TRUE, keep.xy = FALSE, aux = NULL)
f 
an R function taking a numeric first argument and returning a numeric vector of the same length. Returning a nonfinite element will generate an error. 
lower, upper 
the limits of integration. Can be infinite. 
... 
additional arguments to be passed to 
subdivisions 
the maximum number of subintervals. 
rel.tol 
relative accuracy requested. 
abs.tol 
absolute accuracy requested. 
stop.on.error 
logical. If true (the default) an error stops the
function. If false some errors will give a result with a warning in
the 
keep.xy 
unused. For compatibility with S. 
aux 
unused. For compatibility with S. 
Note that arguments after ...
must be matched exactly.
If one or both limits are infinite, the infinite range is mapped onto a finite interval.
For a finite interval, globally adaptive interval subdivision is used in connection with extrapolation by Wynn's Epsilon algorithm, with the basic step being Gauss–Kronrod quadrature.
rel.tol
cannot be less than max(50*.Machine$double.eps,
0.5e28)
if abs.tol <= 0
.
Note that the comments in the C source code in
‘<R>/src/appl/integrate.c’ give more details, particularly about
reasons for failure (internal error code ier >= 1
).
In R versions \le
3.2.x, the first entries of
lower
and upper
were used whereas an error is signalled
now if they are not of length one.
A list of class "integrate"
with components
value 
the final estimate of the integral. 
abs.error 
estimate of the modulus of the absolute error. 
subdivisions 
the number of subintervals produced in the subdivision process. 
message 

call 
the matched call. 
Like all numerical integration routines, these evaluate the function on a finite set of points. If the function is approximately constant (in particular, zero) over nearly all its range it is possible that the result and error estimate may be seriously wrong.
When integrating over infinite intervals do so explicitly, rather than just using a large number as the endpoint. This increases the chance of a correct answer – any function whose integral over an infinite interval is finite must be near zero for most of that interval.
For values at a finite set of points to be a fair reflection of the behaviour of the function elsewhere, the function needs to be wellbehaved, for example differentiable except perhaps for a small number of jumps or integrable singularities.
f
must accept a vector of inputs and produce a vector of function
evaluations at those points. The Vectorize
function
may be helpful to convert f
to this form.
Based on QUADPACK routines dqags
and dqagi
by
R. Piessens and E. deDoncker–Kapenga, available from Netlib.
R. Piessens, E. deDoncker–Kapenga, C. Uberhuber, D. Kahaner (1983) Quadpack: a Subroutine Package for Automatic Integration; Springer Verlag.
integrate(dnorm, 1.96, 1.96)
integrate(dnorm, Inf, Inf)
## a slowlyconvergent integral
integrand < function(x) {1/((x+1)*sqrt(x))}
integrate(integrand, lower = 0, upper = Inf)
## don't do this if you really want the integral from 0 to Inf
integrate(integrand, lower = 0, upper = 10)
integrate(integrand, lower = 0, upper = 100000)
integrate(integrand, lower = 0, upper = 1000000, stop.on.error = FALSE)
## some functions do not handle vector input properly
f < function(x) 2.0
try(integrate(f, 0, 1))
integrate(Vectorize(f), 0, 1) ## correct
integrate(function(x) rep(2.0, length(x)), 0, 1) ## correct
## integrate can fail if misused
integrate(dnorm, 0, 2)
integrate(dnorm, 0, 20)
integrate(dnorm, 0, 200)
integrate(dnorm, 0, 2000)
integrate(dnorm, 0, 20000) ## fails on many systems
integrate(dnorm, 0, Inf) ## works
integrate(dnorm, 0:1, 20) #> error!
## "silently" gave integrate(dnorm, 0, 20) in earlier versions of R