SPSSX Discussion

ANOVA

Classic

List

Threaded

4 messages Options

E. Bernardo

ANOVA

Hi all,

I am using SPSS to run a one-way ANOVA for my three independent groups (samples) with small sample sizes as follows n1=39, n2=4, and n3=2. The testing variable is continuous. Of course, the goal was to test the null hypothesis that the group means are significantly equal.

Some info regarding the data are as follows:
1) The variances of the three groups were statistically insignificant as shown by the levene statistics (LS=1.684, p=.198)
2. skewness of the groups are -.198, -1.441, and -.438
2) group means are 3.9231, 3.9375, and 4.6250 where ANOVA showed insignificant F (F=.849, p=.345)
3) group medians are 4.0000, 4.2500, and 4.6250 where Kruskal Wallis Test is insignificant (chisquare =2.471, p=.291). Note that the Kruskall Wallis Test was ran for your further info only.

I am not confident of the results because of sample size reasons. Any suggestions as to the possible solution are greatly appreciated. Note that there is no more chance of increasing the sample size for each group.

Thank you.
Eins

---------------------------------
Importing contacts has never been easier.
Bring your friends over to Yahoo! Mail today!

=====================
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

E. Bernardo

Re: ANOVA

Hi all,

(My apology for resending).

I am using SPSS to run a one-way ANOVA for my three independent groups (samples) with small sample sizes as follows n1=39, n2=4, and n3=2. The testing variable is continuous. Of course, the goal was to test the null hypothesis that the group means are significantly equal.

Some info regarding the data are as follows:
1) The variances of the three groups were statistically insignificant as shown by the levene statistics (LS=1.684, p=.198)
2. skewness of the groups are -.198, -1.441, and -.438
2) group means are 3.9231, 3.9375, and 4.6250 where ANOVA showed insignificant F (F=.849, p=.345)
3) group medians are 4.0000, 4.2500, and 4.6250 where Kruskal Wallis Test is insignificant (chisquare =2.471, p=.291). Note that the Kruskall Wallis Test was ran for your further info only.

I am not confident of the results because of sample size reasons. Any suggestions as to the possible solution are greatly appreciated. Note that there is no more chance of increasing the sample size for each group.

Thank you.
Eins

---------------------------------
Catch Emoticarolers in the act in this Christmas!
Spread holiday cheers to your friends and loved ones via Yahoo! Messenger today!

=====================
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

Art Kendall

Re: ANOVA

The term "significantly equal" is problematic.

The p of .345 says there is not sufficient evidence to assert that some
pair of means is inconsistent with the h that the means differ only
random processes.

You tested the H that some pair of means if guilty of being different by
a reasonable standard. The *guilt was not proven *and therefore you
have to act as if they were innocent (the same) in terms of theory,
practice, or policy.

Art Kendall
Social Research Consultants

Eins Bernardo wrote:

> Hi all,
>
> (My apology for resending).
>
> I am using SPSS to run a one-way ANOVA for my three independent groups (samples) with small sample sizes as follows n1=39, n2=4, and n3=2. The testing variable is continuous. Of course, the goal was to test the null hypothesis that the group means are significantly equal.
>
> Some info regarding the data are as follows:
> 1) The variances of the three groups were statistically insignificant as shown by the levene statistics (LS=1.684, p=.198)
> 2. skewness of the groups are -.198, -1.441, and -.438
> 2) group means are 3.9231, 3.9375, and 4.6250 where ANOVA showed insignificant F (F=.849, p=.345)
> 3) group medians are 4.0000, 4.2500, and 4.6250 where Kruskal Wallis Test is insignificant (chisquare =2.471, p=.291). Note that the Kruskall Wallis Test was ran for your further info only.
>
> I am not confident of the results because of sample size reasons. Any suggestions as to the possible solution are greatly appreciated. Note that there is no more chance of increasing the sample size for each group.
>
> Thank you.
> Eins
>
>
>
>
> ---------------------------------
> Catch Emoticarolers in the act in this Christmas!
> Spread holiday cheers to your friends and loved ones via Yahoo! Messenger today!
>
> =====================
> 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
>
>
>

=====================
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

Art Kendall
Social Research Consultants

Albert-Jan Roskam

Re: ANOVA

Hi,

A significantly happy 2009 (p < .0001) to you all!

Re: the example below, I was wondering to what extent the statistical assumptions of ANOVA are met? Did the Levene Statistic indicate that variance is homogenous? Were the differential skewness measures an indication of non-normality? If the assumptions were violated: (what) does it matter? I am just wondering, because to me it seems that assumptions are often 'assumed'.

Cheers!!
Albert-Jan

--- On Sun, 1/4/09, Art Kendall <[hidden email]> wrote:

> From: Art Kendall <[hidden email]>
> Subject: Re: ANOVA
> To: [hidden email]
> Date: Sunday, January 4, 2009, 5:56 PM
> The term "significantly equal" is problematic.
>
> The p of .345 says there is not sufficient evidence to
> assert that some
> pair of means is inconsistent with the h that the means
> differ only
> random processes.
>
> You tested the H that some pair of means if guilty of being
> different by
> a reasonable standard. The *guilt was not proven *and
> therefore you
> have to act as if they were innocent (the same) in terms of
> theory,
> practice, or policy.
>
> Art Kendall
> Social Research Consultants
>
>
> Eins Bernardo wrote:
> > Hi all,
> >
> > (My apology for resending).
> >
> > I am using SPSS to run a one-way ANOVA for my three
> independent groups (samples) with small sample sizes as
> follows n1=39, n2=4, and n3=2. The testing variable is
> continuous. Of course, the goal was to test the null
> hypothesis that the group means are significantly equal.
> >
> > Some info regarding the data are as follows:
> > 1) The variances of the three groups were
> statistically insignificant as shown by the levene
> statistics (LS=1.684, p=.198)
> > 2. skewness of the groups are -.198, -1.441, and
> -.438
> > 2) group means are 3.9231, 3.9375, and 4.6250 where
> ANOVA showed insignificant F (F=.849, p=.345)
> > 3) group medians are 4.0000, 4.2500, and 4.6250
> where Kruskal Wallis Test is insignificant (chisquare
> =2.471, p=.291). Note that the Kruskall Wallis Test was ran
> for your further info only.
> >
> > I am not confident of the results because of sample
> size reasons. Any suggestions as to the possible solution
> are greatly appreciated. Note that there is no more chance
> of increasing the sample size for each group.
> >
> > Thank you.
> > Eins
> >
> >
> >
> >
> > ---------------------------------
> > Catch Emoticarolers in the act in this Christmas!
> > Spread holiday cheers to your friends and loved ones
> via Yahoo! Messenger today!
> >
> > =====================
> > 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
> >
> >
> >
>
> =====================
> 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