Correcting for Unbalanced Design

[Courtney M. Cronley]

Hello,

I have a question. I am trying to look at group differences among five groups. The problem is that the groups are unequally divided:

G1 = 128
G2 = 31
G3 = 113
G4 = 157
G5 = 30

I'm running ANOVAs, binary regression, and negative binomial regression models. Is there any way to control for the unbalanced design?

Thanks,

Courtney

Courtney Cronley, Ph.D.
Postdoctoral Associate
Center of Alcohol Studies
Rutgers University
[hidden email]

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

There is no requirement that the sample sizes be equal.  ANOVA is more robust to heterogeneity of variance when the sample sizes are equal (or nearly so), but equal sample sizes are not required.  If there is enough heterogeneity of variance to be concerning, you could always use the Welch or Brown-Forsythe version of the F-test, which does not require equal variances.

Courtney M. Cronley wrote

Hello,

I have a question. I am trying to look at group differences among five groups. The problem is that the groups are unequally divided:

G1 = 128
G2 = 31
G3 = 113
G4 = 157
G5 = 30

I'm running ANOVAs, binary regression, and negative binomial regression models. Is there any way to control for the unbalanced design?

Thanks,

Courtney

Courtney Cronley, Ph.D.
Postdoctoral Associate
Center of Alcohol Studies
Rutgers University
[e-mail]

=====================
To manage your subscription to SPSSX-L, send a message to
[e-mail] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Bruce Weaver]

I should add that I assumed you are talking about one-way ANOVA.  In factorial ANOVA, lack of balance causes problems if you don't take measures to deal with it.  The most common method for dealing with it is use of Type III sums of squares.

Bruce Weaver wrote

There is no requirement that the sample sizes be equal.  ANOVA is more robust to heterogeneity of variance when the sample sizes are equal (or nearly so), but equal sample sizes are not required.  If there is enough heterogeneity of variance to be concerning, you could always use the Welch or Brown-Forsythe version of the F-test, which does not require equal variances.

Courtney M. Cronley wrote

Hello,

I have a question. I am trying to look at group differences among five groups. The problem is that the groups are unequally divided:

G1 = 128
G2 = 31
G3 = 113
G4 = 157
G5 = 30

I'm running ANOVAs, binary regression, and negative binomial regression models. Is there any way to control for the unbalanced design?

Thanks,

Courtney

Courtney Cronley, Ph.D.
Postdoctoral Associate
Center of Alcohol Studies
Rutgers University
[e-mail]

=====================
To manage your subscription to SPSSX-L, send a message to
[e-mail] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD