alternative to anova

What is an alternative statistical test to anova. I have a large varience between three groups (control and two experimental). How can I determine if there is a significant difference between the means (p-value). My n=12. Thanks in advance.


Fortran must die
What is the dependent variable? That is how is it measured.

And why don't you want to run ANOVA (which will influence what method you replace it with since alternatives like T test are very similar in assumption).
Dependent variable is transcriptional activity for a reporter assay. I ran anova but the variance for my groups are very different. I read that you should not use anova if the variances between groups not similar.

Varience of data:
Control = <1
Group 1 = 32
Group 2 = 19


Ambassador to the humans
When you say n = 12 do you mean you have 12 subjects in each group or 12 subjects total (4 in each group?)
I have data from 12 separate experiments. Each experiment contains a single data value for each of 3 groups (1 control and 2 experimental). So I have a total of 36 data points; 12 data points for each group.


Fortran must die
The inequality of the population variances can be assessed by examination of the relative size of the sample variances, either informally (including graphically), or by a robust variance test such as Levene's test. (Bartlett's test is even more sensitive to nonnormality than the one-way ANOVA's F test, and thus should not be used for such testing.) The effect of inequality of variances is mitigated when the sample sizes are equal: The F test is fairly robust against inequality of variances if the sample sizes are equal, although the chance increases of incorrectly reporting a significant difference in the means when none exists. This chance of incorrectly rejecting the null hypothesis is greater when the population variances are very different from each other, particularly if there is one sample variance very much larger than the others.
The effect of inequality of the variances is most severe when the sample sizes are unequal. If the larger samples are associated with the populations with the larger variances, then the F statistic will tend to be smaller than it should be, reducing the chance that the test will correctly identify a significant difference between the means (i.e., making the test conservative).
How different are your populations (your group) variances from each other? And how similar are their sample sizes?

Did you run the Levene test and it showed they were in fact signficantly unequal?


TS Contributor
I have data from 12 separate experiments. Each experiment contains a single data value for each of 3 groups (1 control and 2 experimental).
Are the groups really independent? Couldn't this be a repeated-measures design
(12 "subjects", 3 levels)? What is the nature of the intervention/experiment
and of the data points? Could you explain it in a bit more detail?

Kind regards



Ambassador to the humans
Well if it's 12 different experiments I don't think there would be much of a problem with independence
The Levene test certainly sounds like a good start. If I find that my variance is significantly different where do i go from there? What is the right test?

It might be appropriate to mention that variation between my control and experimental groups is large because of the sensitivity of the machine that I collect my data from. The ratio of the means between the two experimental groups is usually consistent. If this makes sense or is helpful in your assisting me great. If not forget that I mentioned it.

I listed the variances in post 2. Sample sizes are the same for all groups.


Fortran must die
If you have unequal variances I would imagine that you would have heteroskedacity so many forms of regression would have problems as would t test. Weighted least squares or a transformation of the data might work, although I suspect some form of ANOVA is still your best bet. But it would be a more complex design than I work with.