Help with bootci

[Ruth Watson]

Hi there,
I have a very simple problem, which I am hoping will have a simple solution. I am trying to use the bootci function but I receive the following error message:

??? Error using ==> jackknife
Too many input arguments.

Error in ==> bootci>bootbca at 216
jstat = jackknife(bootfun,varargin{:});

Error in ==> bootci at 127
ci = bootbca(nboot,fun,alpha,data{:});

even when I try something very simple such as:
data=[1 2 3 4 5 6 7 8 9 10];
ci = bootci(100,@mean,data);

I am sure that I have previously used this function without problems in an earlier version of Matlab. I am currently running version 7.8.0.347 (R2009a). I tried using a different type of bootci. eg

ci = bootci(100,@mean,data,'type','cper') but then received a different error message (maybe I am not specifying the type correctly?):

??? Error using ==> bootci at 106
Error evaluating BOOTFUN function 'mean'.
Caused by:
Error using ==> mean
Too many input arguments.

Thanks for any help,
Ruth

[Han Oostdijk]

>
> even when I try something very simple such as:
> data=[1 2 3 4 5 6 7 8 9 10];
> ci = bootci(100,@mean,data);
>

Ruth,

the code above works for me.
You are sure you did not overwrite function mean?

Regards Han

[Jackson]

Greetings,

I'm getting the same problem as Ruth:

-----------------------------------------------
>> mean(rand(10,1))
ans =
0.4233

>> bootci(2000,@mean,rand(10,1))

??? Error using ==> jackknife
Too many input arguments.

Error in ==> bootci>bootbca at 260
jstat = jackknife(bootfun,varargin{:},'Options',bootstrpOptions);

Error in ==> bootci at 170
[ci,bootstat] = bootbca(obsstat,nboot,fun,alpha,weights,bootstrpOptions,data{:});
-----------------------------------------------

I'm running R2010b on a 64-bit intel mac with Snow Leopard.
The function 'mean' is not overridden. The same thing happens using 'median', 'std', 'var', and so forth.

This looks very much like a Mathworks bug.

Regards,

Jack

[Tom Lane]

> ci = bootci(100,@mean,data,'type','cper') but then received a different
> error message (maybe I am not specifying the type correctly?):

The trick here is to separate the arguments to bootci from the arguments
intended to be passed along to mean. Try this:

ci = bootci(100, {@mean,data}, 'type','cper')

Note that I've grouped the mean function with its arguments by enclosing
them together in a cell array.

>>> bootci(2000,@mean,rand(10,1))
> ??? Error using ==> jackknife
> Too many input arguments.

This one I am unable to reproduce. It appears to be claiming that you've
passed more arguments to jackknife than it is declared to handle. But it
should be declared to handle a variable number. Would you mind trying this?

>> dbtype jackknife 1

1 function jackstat = jackknife(jackfun,varargin)

>> which jackknife
...\matlab\toolbox\stats\stats\jackknife.m

This shows that I'm invoking the Statistics Toolbox jackknife function, and
it is defined to accept a variable number of inputs.

-- Tom

[Ruth Watson]

Thanks to everyone for the help! The problem is now fixed, as detailed below:

> ci = bootci(100, {@mean,data}, 'type','cper')

Yes, thanks, this now works for me.

> >>> bootci(2000,@mean,rand(10,1))
> > ??? Error using ==> jackknife
> > Too many input arguments.

> >> dbtype jackknife 1
>
> 1 function jackstat = jackknife(jackfun,varargin)
>
> >> which jackknife
> ...\matlab\toolbox\stats\stats\jackknife.m

I tried the above and found that I was calling a different jackknife function (from the Chronux spectral analysis toolbox I downloaded recently). I renamed that function and now >> bootci(2000,@mean,rand(10,1)) works just fine.

Sorry for not being able to solve this problem myself (doh!) and thanks again for the prompt help.

Ruth