turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- possible to use ANOVA to compare medians?

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

04-01-2012 12:14 AM

Well, there might be such a thing as a 'stupid question'. I measured a variable having an underlying continuous distribution (distance between two animals: mother and offspring) and calculated median values for these pairs. I mostly used median because distance was divided into some categories that are uneven (larger categories as the distance gets longer).

The distribution of distances ended up being relatively normal with equal variance among groups. Is it possible to use an ANOVA (parametric) to compare these medians rather than a non-parametric test such as Kruskal-Wallis? The ANOVA has the advantage of being able to include two variables at once.

Thanks in advance!

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

04-01-2012 01:51 AM

My opinion is no.

You can do ANOVA because F statistical estimator .

I don't think you can figure out a way to make a F by using Median .

Ksharp

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

04-01-2012 03:11 PM

If I understand correctly, you are proposing to compare the median distance between mother and offspring between groups of animals with multiway ANOVA. This distance would define the area around the mother inside which you are likely to find its offspring half the time. You could have used any quantile of the distances. I say: go ahead with the ANOVA. Just never call your measurements mean distances and don't forget to test for interactions.

PG

PG