bootci

Bootstrap confidence interval

Syntax

ci = bootci(nboot,bootfun,d)

ci = bootci(nboot,bootfun,d1,...,dN)

ci = bootci(nboot,{bootfun,d},Name,Value)

ci = bootci(nboot,{bootfun,d1,...,dN},Name,Value)

[ci,bootstat] = bootci(___)

Description

ci = bootci(nboot,bootfun,d) computes a 95% bootstrap confidence interval for each statistic computed by the function bootfun. The bootci function uses nboot bootstrap samples in its computation, and creates each bootstrap sample by sampling with replacement from the rows of d.

ci = bootci(nboot,bootfun,d1,...,dN) creates each bootstrap sample by sampling with replacement from the rows of the nonscalar data arguments in d1,...,dN. These nonscalar arguments must have the same number of rows. The bootci function passes the samples of nonscalar data and the unchanged scalar data arguments in d1,...,dN to bootfun.

ci = bootci(nboot,{bootfun,d},Name,Value) specifies options using one or more name-value arguments. For example, you can change the type of confidence interval by specifying the 'Type' name-value argument.

Note that you must pass the bootfun and d arguments to bootci as a single cell array.

ci = bootci(nboot,{bootfun,d1,...,dN},Name,Value) specifies options using one or more name-value arguments. For example, you can change the significance level of the confidence interval by specifying the 'Alpha' name-value argument.

Note that you must pass the bootfun and d1,...,dN arguments to bootci as a single cell array.

[ci,bootstat] = bootci(___) also returns the bootstrapped statistic computed for each of the nboot bootstrap replicate samples, using any of the input argument combinations in the previous syntaxes. Each row of bootstat contains the results of applying bootfun to one bootstrap sample.

Examples

Bootstrap Confidence Interval

Compute the confidence interval for the capability index in statistical process control.

Generate 30 random numbers from the normal distribution with mean 1 and standard deviation 1.

rng('default') % For reproducibility
y = normrnd(1,1,30,1);

Specify the lower and upper specification limits of the process. Define the capability index.

LSL = -3;
USL = 3;
capable = @(x)(USL-LSL)./(6*std(x));

Compute the 95% confidence interval for the capability index by using 2000 bootstrap samples. By default, bootci uses the bias corrected and accelerated percentile method to construct the confidence interval.

ci = bootci(2000,capable,y)

ci = 2×1

    0.5937
    0.9900

Compute the studentized confidence interval for the capability index.

sci = bootci(2000,{capable,y},'Type','student')

sci = 2×1

    0.5193
    0.9930

Bootstrap Confidence Intervals for Nonlinear Regression Model Coefficients

Compute bootstrap confidence intervals for the coefficients of a nonlinear regression model. The technique used in this example involves bootstrapping the predictor and response values, and assumes that the predictor variable is random. For a technique that assumes the predictor variable is fixed and bootstraps the residuals, see Bootstrap Confidence Intervals for Linear Regression Model Coefficients.

Note: This example uses nlinfit, which is useful when you only need the coefficient estimates or residuals of a nonlinear regression model and you need to repeat fitting a model multiple times, as in the case of bootstrapping. If you need to investigate a fitted regression model further, create a nonlinear regression model object by using fitnlm. You can create confidence intervals for the coefficients of the resulting model by using the coefCI object function, although this function does not use bootstrapping.

Generate data from the nonlinear regression model $$y=b_1+b_2\cdot exp(-b_{3}x)+ϵ$$, where $$b_1=1$$, $$b_2=3$$, and $$b_3=2$$ are coefficients; the predictor variable x is exponentially distributed with mean 2; and the error term $$ϵ$$ is normally distributed with mean 0 and standard deviation 0.1.

modelfun = @(b,x)(b(1)+b(2)*exp(-b(3)*x));

rng('default') % For reproducibility
b = [1;3;2];
x = exprnd(2,100,1);
y = modelfun(b,x) + normrnd(0,0.1,100,1);

Create a function handle for the nonlinear regression model that uses the initial values in beta0.

beta0 = [2;2;2];
beta = @(predictor,response)nlinfit(predictor,response,modelfun,beta0)

beta = function_handle with value:
    @(predictor,response)nlinfit(predictor,response,modelfun,beta0)

Compute the 95% bootstrap confidence intervals for the coefficients of the nonlinear regression model. Create the bootstrap samples from the generated data x and y.

ci = bootci(1000,beta,x,y)

ci = 2×3

    0.9821    2.9552    2.0180
    1.0410    3.1623    2.2695

The first two confidence intervals include the true coefficient values $$b_1=1$$ and $$b_2=3$$, respectively. However, the third confidence interval does not include the true coefficient value $$b_3=2$$.

Now compute the 99% bootstrap confidence intervals for the model coefficients.

newci = bootci(1000,{beta,x,y},'Alpha',0.01)

newci = 2×3

    0.9730    2.9112    1.9562
    1.0469    3.1876    2.3133

All three confidence intervals include the true coefficient values.

Bootstrap Confidence Intervals for Linear Regression Model Coefficients

Compute bootstrap confidence intervals for the coefficients of a linear regression model. The technique used in this example involves bootstrapping the residuals and assumes that the predictor variable is fixed. For a technique that assumes the predictor variable is random and bootstraps the predictor and response values, see Bootstrap Confidence Intervals for Nonlinear Regression Model Coefficients.

Note: This example uses regress, which is useful when you only need the coefficient estimates or residuals of a regression model and you need to repeat fitting a model multiple times, as in the case of bootstrapping. If you need to investigate a fitted regression model further, create a linear regression model object by using fitlm. You can create confidence intervals for the coefficients of the resulting model by using the coefCI object function, although this function does not use bootstrapping.

Load the sample data.

load hald

Perform a linear regression and compute the residuals.

x = [ones(size(heat)),ingredients];
y = heat;
b = regress(y,x);
yfit = x*b;
resid = y - yfit;

Compute the 95% bootstrap confidence intervals for the coefficients of the linear regression model. Create the bootstrap samples from the residuals. Use normal approximated intervals with bootstrapped bias and standard error by specifying 'Type','normal'. You cannot use the default confidence interval type in this case.

ci = bootci(1000,{@(bootr)regress(yfit+bootr,x),resid}, ...
    'Type','normal')

ci = 2×5

  -47.7130    0.3916   -0.6298   -1.0697   -1.2604
  172.4899    2.7202    1.6495    1.2778    0.9704

Plot the estimated coefficients b, omitting the intercept term, and display error bars showing the coefficient confidence intervals.

slopes = b(2:end)';
lowerBarLengths = slopes-ci(1,2:end);
upperBarLengths = ci(2,2:end)-slopes;
errorbar(1:4,slopes,lowerBarLengths,upperBarLengths)
xlim([0 5])
title('Coefficient Confidence Intervals')

Only the first nonintercept coefficient is significantly different from 0.

Confidence Intervals for Multiple Statistics

Compute the mean and standard deviation of 100 bootstrap samples. Find the 95% confidence interval for each statistic.

Generate 100 random numbers from the exponential distribution with mean 5.

rng('default') % For reproducibility
y = exprnd(5,100,1);

Draw 100 bootstrap samples from the vector y. For each bootstrap sample, compute the mean and standard deviation. Find the 95% bootstrap confidence interval for the mean and standard deviation.

[ci,bootstat] = bootci(100,@(x)[mean(x) std(x)],y);

ci(:,1) contains the lower and upper bounds of the mean confidence interval, and c(:,2) contains the lower and upper bounds of the standard deviation confidence interval. Each row of bootstat contains the mean and standard deviation of a bootstrap sample.

Plot the mean and standard deviation of each bootstrap sample as a point. Plot the lower and upper bounds of the mean confidence interval as dotted vertical lines, and plot the lower and upper bounds of the standard deviation confidence interval as dotted horizontal lines.

plot(bootstat(:,1),bootstat(:,2),'o')
xline(ci(1,1),':')
xline(ci(2,1),':')
yline(ci(1,2),':')
yline(ci(2,2),':')
xlabel('Mean')
ylabel('Standard Deviation')

Input Arguments

nboot — Number of bootstrap samples

positive integer scalar

Number of bootstrap samples to draw, specified as a positive integer scalar. To create each bootstrap sample, bootci randomly selects with replacement n out of the n rows of nonscalar data in d or d1,...,dN.

Example: 100

Data Types: single | double

bootfun — Function to apply to each sample

function handle

Function to apply to each sample, specified as a function handle. The function can be a custom or built-in function. You must specify bootfun with the @ symbol.

Example: @mean

Data Types: function_handle

d — Data to sample from

column vector | matrix

Data to sample from, specified as a column vector or matrix. The n rows of d correspond to observations. When you use multiple data input arguments d1,...,dN, you can specify some arguments as scalar values, but all nonscalar arguments must have the same number of rows.

If you use a single vector argument d, you can specify it as a row vector. bootci then samples from the elements of the vector.

Data Types: single | double

Name-Value Arguments

Specify optional pairs of arguments as Name1=Value1,...,NameN=ValueN, where Name is the argument name and Value is the corresponding value. Name-value arguments must appear after other arguments, but the order of the pairs does not matter.

Before R2021a, use commas to separate each name and value, and enclose Name in quotes.

Example: bootci(100,{@mean,1:6'},'Alpha',0.1) specifies to draw 100 bootstrap samples from the values 1 through 6, take the mean of each sample, and then compute the 90% confidence interval for the sample mean.

Alpha — Significance level

0.05 (default) | positive scalar in (0,1)

Significance level, specified as a positive scalar between 0 and 1. bootci computes the 100*(1-Alpha) bootstrap confidence interval of each statistic defined by the function bootfun.

Example: 'Alpha',0.01

Data Types: single | double

Type — Confidence interval type

'bca' (default) | 'norm' | 'per' | 'cper' | 'stud' | ...

Confidence interval type, specified as one of the values in this table.

Value	Description
'norm' or 'normal'	Normal approximated interval with bootstrapped bias and standard error [1]
'per' or 'percentile'	Basic percentile method
'cper' or 'corrected percentile'	Bias corrected percentile method [2]
'bca'	Bias corrected and accelerated percentile method [3], [4]. This method involves a z0 factor computed using the proportion of bootstrap values that are less than the original sample value. To produce reasonable results when the sample is lumpy, the software computes z0 by including half of the bootstrap values that are the same as the original sample value.
'stud' or 'student'	Studentized confidence interval [3]

Example: 'Type','student'

NBootStd — Number of bootstrap samples for studentized standard error estimate

100 (default) | positive integer scalar

Number of bootstrap samples for the studentized standard error estimate, specified as a positive integer scalar.

bootci computes the studentized bootstrap confidence interval of the statistic defined by the function bootfun, and estimates the standard error of the bootstrap statistics by using NBootStd bootstrap data samples.

Note

To use this name-value argument, the Type value must be 'stud' or 'student'. Specify either NBootStd or StdErr, but not both.

Example: 'NBootStd',50

Data Types: single | double

StdErr — Function used to compute studentized standard error estimate

function handle

Function used to compute the studentized standard error estimate, specified as a function handle.

bootci computes the studentized bootstrap confidence interval of the statistic defined by the function bootfun, and estimates the standard error of the bootstrap statistics by using the function StdErr. The StdErr function must take the same arguments as bootfun and return the standard error of the statistic computed by bootfun.

Note

To use this name-value argument, the Type value must be 'stud' or 'student'. Specify either NBootStd or StdErr, but not both.

Example: 'StdErr',@std

Data Types: function_handle

Weights — Observation weights

ones(n,1)/n (default) | nonnegative vector

Observation weights, specified as a nonnegative vector with at least one positive element. The number of elements in Weights must be equal to the number of rows n in the data d or d1,...,dN. To obtain one bootstrap sample, bootci randomly selects with replacement n out of n rows of data using these weights as multinomial sampling probabilities.

Data Types: single | double

Options — Options for computing iterations in parallel and setting random numbers

structure

Options for computing bootstrap iterations in parallel and setting random numbers during the bootstrap sampling, specified as a structure. Create the Options structure with statset. This table lists the option fields and their values.

Field Name	Value	Default
UseParallel	Set this value to true to compute bootstrap iterations in parallel.	false
UseSubstreams	Set this value to true to run computations in parallel in a reproducible manner. To compute reproducibly, set Streams to a type that allows substreams: 'mlfg6331_64' or 'mrg32k3a'.	false
Streams	Specify this value as a RandStream object or cell array of such objects. Use a single object except when the UseParallel value is true and the UseSubstreams value is false. In that case, use a cell array that has the same size as the parallel pool.	If you do not specify Streams, then bootci uses the default stream or streams.

Note

You need Parallel Computing Toolbox™ to run computations in parallel.

Example: 'Options',statset('UseParallel',true)

Data Types: struct

Output Arguments

ci — Confidence interval bounds

vector with two rows | matrix with two rows | multidimensional array with two rows

Confidence interval bounds, returned as a vector, matrix, or multidimensional array with two rows.

• If bootfun returns a scalar, then ci is a vector containing the lower and upper bounds of the confidence interval.

• If bootfun returns a vector of length m, then ci is a matrix of size 2-by-m, where ci(1,:) are lower bounds and ci(2,:) are upper bounds.

• If bootfun returns a multidimensional array, then ci is an array, where ci(1,:,...) is an array of lower bounds and ci(2,:,...) is an array of upper bounds.

bootstat — Bootstrap statistics

column vector | matrix

Bootstrap statistics, returned as a column vector or matrix with nboot rows. The ith row of bootstat corresponds to the results of applying bootfun to the ith bootstrap sample. If bootfun returns a matrix or array, then the bootci function first converts this output to a row vector before storing it in bootstat.

References

[1] Davison, A. C., and D. V. Hinkley. Bootstrap Methods and Their Applications. Cambridge University Press, 1997.

[2] Efron, Bradley. The Jackknife, the Bootstrap and Other Resampling Plans. Philadelphia: The Society for Industrial and Applied Mathematics, 1982.

[3] DiCiccio, Thomas J., and Bradley Efron. “Bootstrap Confidence Intervals.” Statistical Science 11, no. 3 (1996): 189–228.

[4] Efron, Bradley, and Robert J. Tibshirani. An Introduction to the Bootstrap. New York: Chapman & Hall, 1993.

Extended Capabilities

Automatic Parallel Support

Accelerate code by automatically running computation in parallel using Parallel Computing Toolbox™.

To run in parallel, specify the 'Options' name-value argument in the call to this function and set the 'UseParallel' field of the options structure to true using statset.

For example: 'Options',statset('UseParallel',true)

For more information about parallel computing, see Run MATLAB Functions with Automatic Parallel Support (Parallel Computing Toolbox).

Version History

Introduced in R2006a