MATLAB Central Newsreader - Unbalanced ANOVA

Unbalanced ANOVA

Suman Seer — Thu, 14 Aug 2008 17:42:02 -0400

Hello Everyone,

I have been using anova1 till now to check for equality of
means between groups with same number of elements. But now
my data has unequal number of elements in groups and I
believe this is the case of Unbalanced ANOVA. I havent been
able to figure out a way or function to handle this case?
I would appreciate any help with this.

Thanks
Suman

Re: Unbalanced ANOVA

Donn Shull — Thu, 14 Aug 2008 18:33:01 -0400

"Suman Seer" <supernova5271@yahoo.com> wrote in message
<g81qpa$6tk$1@fred.mathworks.com>...
> Hello Everyone,
>
> I have been using anova1 till now to check for equality of
> means between groups with same number of elements. But now
> my data has unequal number of elements in groups and I
> believe this is the case of Unbalanced ANOVA. I havent
been
> able to figure out a way or function to handle this case?
> I would appreciate any help with this.
>
> Thanks
> Suman

group1 = 1:3;
group2 = 2;

groups = [group1, group2];
names = {'gp1','gp1','gp1','gp2'};

[p,anovatab,stats] = anova1(groups, names);

Re: Unbalanced ANOVA

Suman Seer — Thu, 14 Aug 2008 20:42:01 -0400

Hi Donn,

Thanks for your reply. This is how I have been using
anova1 before but my question is, since the groups have
unequal elements, is the anova1 test still valid or does
one have to make any changes?

Thanks.
Hussain

"Donn Shull" <donn.shull.no_spam@aetoolbox.com> wrote in
message <g81tot$co6$1@fred.mathworks.com>...
> "Suman Seer" <supernova5271@yahoo.com> wrote in message
> <g81qpa$6tk$1@fred.mathworks.com>...
> > Hello Everyone,
> >
> > I have been using anova1 till now to check for
equality of
> > means between groups with same number of elements. But
now
> > my data has unequal number of elements in groups and I
> > believe this is the case of Unbalanced ANOVA. I havent
> been
> > able to figure out a way or function to handle this
case?
> > I would appreciate any help with this.
> >
> > Thanks
> > Suman
>
> group1 = 1:3;
> group2 = 2;
>
> groups = [group1, group2];
> names = {'gp1','gp1','gp1','gp2'};
>
> [p,anovatab,stats] = anova1(groups, names);
>
>

Re: Unbalanced ANOVA

Donn Shull — Thu, 14 Aug 2008 21:17:02 -0400

"Suman Seer" <supernova5271@yahoo.com> wrote in message
<g825ap$mkf$1@fred.mathworks.com>...
> Hi Donn,
>
> Thanks for your reply. This is how I have been using
> anova1 before but my question is, since the groups have
> unequal elements, is the anova1 test still valid or does
> one have to make any changes?
>
> Thanks.
> Hussain
>
>
>
> "Donn Shull" <donn.shull.no_spam@aetoolbox.com> wrote in
> message <g81tot$co6$1@fred.mathworks.com>...
> > "Suman Seer" <supernova5271@yahoo.com> wrote in message
> > <g81qpa$6tk$1@fred.mathworks.com>...
> > > Hello Everyone,
> > >
> > > I have been using anova1 till now to check for
> equality of
> > > means between groups with same number of elements.
But
> now
> > > my data has unequal number of elements in groups and I
> > > believe this is the case of Unbalanced ANOVA. I
havent
> > been
> > > able to figure out a way or function to handle this
> case?
> > > I would appreciate any help with this.
> > >
> > > Thanks
> > > Suman
> >
> > group1 = 1:3;
> > group2 = 2;
> >
> > groups = [group1, group2];
> > names = {'gp1','gp1','gp1','gp2'};
> >
> > [p,anovatab,stats] = anova1(groups, names);
> >
> >
>

Hi Hussain,

That question is beyond my area expertise. Perhaps someone
else can answer that. My guess is that since specifically
addresses how to use anova1 with unequal groups that it
would be valid. Anyone else?

Donn

Re: Unbalanced ANOVA

Tom Lane — Fri, 15 Aug 2008 01:43:37 -0400

> Thanks for your reply. This is how I have been using
> anova1 before but my question is, since the groups have
> unequal elements, is the anova1 test still valid or does
> one have to make any changes?

Suman, anova1 accepts unbalanced data and handles that situation properly.

In contrast, you can't use unbalanced data with anova2 -- the input format
just doesn't permit it. Fortunately you can use anovan for unbalanced data
with two or more grouping variables.

-- Tom

Re: Unbalanced ANOVA

Farah — Fri, 29 May 2009 11:41:01 -0400

Hi,

I'm new to MatLab and have come against the same problem as Suman. Only I have about 150 different groups, and the number of elements in each varies from 30 to 7000, with no two groups having the same number of elements necessarily. Also the values of the elements themselves are completely unrelated to each other (they are microarray probe log2ratios).
Anova1 is the test I need to do, but I don't quite understand how to apply the function for my case. Can anyone please show me some pointers? I've read Don's reply to Suman but I'm not clear on what it means.

Thanks!

"Tom Lane" <tlane@mathworks.com> wrote in message <g82n0a$eqr$1@fred.mathworks.com>...
> > Thanks for your reply. This is how I have been using
> > anova1 before but my question is, since the groups have
> > unequal elements, is the anova1 test still valid or does
> > one have to make any changes?
>
> Suman, anova1 accepts unbalanced data and handles that situation properly.
>
> In contrast, you can't use unbalanced data with anova2 -- the input format
> just doesn't permit it. Fortunately you can use anovan for unbalanced data
> with two or more grouping variables.
>
> -- Tom
>
>