1
$\begingroup$

Hello I'm running anova with R and I am wondering what the differce is between multiway anova and single anova. I know single anova gives a different answer than multiway anova but I dont know how to quantify multiway anova. Could someone explain in words the difference between:

  • base~col1
  • base~col1+col2
  • base~col1*col2

http://www.gardenersown.co.uk/Education/Lectures/R/anova.htm#anova_1

How do I interpret the results of col1+col2 and col1*col2 because col2 is giving a P(>F) of 0.473 on single anova but P(>F) 4.21e-07 in multiway anova. col2 does correlate well with base having only a 2% coloration for 711 individuals. According to http://vassarstats.net/rsig.html 2% coloration isn't significant. Thank you

edit: i know my col1 and col2 have some interactions is that causing the low multiple anova p value?

i am using

aov.ex2 <- aov(base~col1+col2,data=data1) 
summary(aov.ex2)
aov.ex2 <- aov(base~col1,data=data1) 
summary(aov.ex2)
aov.ex2 <- aov(base~col1,data=data1) 
summary(aov.ex2)

and the results are not simular

$\endgroup$

1 Answer 1

4
$\begingroup$

Simply, single ANOVA tests between-groups for a single independent variable (col1) for a response variable (base). Multiway ANOVA tests between-groups for two or more independent variables (col1 and col2). The choice to use one or the other depends on your experimental design.

base~col1 is testing a single "main" effect

base~col1+col2 would be testing two "main" effects only

base~col1*col2 would be testing two "main" effects along with the "interaction" effect between independent variables.

Some more info about two-way ANOVA

Edit: Given your updated post:

I suspect you made a mistake in the code you gave as example, but here are some explanations:

aov.ex2 <- aov(base~col1+col2,data=data1) 
summary(aov.ex2)

This would test for only main effects of col1 and main effects of col2. In other words, are there any differences between groups for col1 or col2?

aov.ex2 <- aov(base~col1,data=data1) 
summary(aov.ex2)

This would test for only the main effect of col1. In other words, are there any differences between groups for col1?

It seems like your basic problem is that with the inclusion of a second independent variable you are getting different results. This post may be of relevance.

$\endgroup$
12
  • $\begingroup$ ok that makes some sense what would you say to col2 not corolating to base but having a low p vaule. what does that mean? i really need to know why this is happening and why there is no p value for it individually but there is a p vale for it coroporately $\endgroup$
    – user11279
    May 15, 2012 at 15:55
  • $\begingroup$ @caseyr547 can you give us an example of the code you're using (I'm assuming aov() from your previous link) and output? It's unclear what effects you're referring to. $\endgroup$ May 15, 2012 at 16:47
  • $\begingroup$ i updated the question adding some code :) $\endgroup$
    – user11279
    May 15, 2012 at 17:07
  • $\begingroup$ the edit you just made is missing information :) I'm excited to see what you have to say $\endgroup$
    – user11279
    May 15, 2012 at 19:05
  • $\begingroup$ Should be updated now. Could you give us a better idea of the dataset? How many groups are you comparing between? Is the data parametric? $\endgroup$ May 15, 2012 at 19:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.