We are currently handling some results and performing some ANOVA tests on human search performance in virtual environments using different sound systems for audio rendering. We measure the time it takes a participant to find a sound in a virtual environment, when the sound is played through different systems.

Our independent variable is "sound system" and our dependent variable is "time". All subjects have searched for a sound using all types of available systems (Within). The test is a complete test.

The problem is that when we perform an One-Way ANOVA, to see whether there is a significant difference between the time spent searching for a sound using a specific sound system, the resulting P-Value is 0.064.

But using a Two-Way Anova the P-value turns into 0.0217.

The One-Way Anova:

enter image description here

The Two-Way Anova:

enter image description here

The one you should look at is (asfactor(soundSystem)).

We expected that the two values were the same, although they are not. What are we missing here?

The tests have been performed in R.


When you add variables to an ANOVA (or any regression model) you are controlling for it. You added two variables (amount of distractors and an interaction) so, of course all the values changed. The only way they could not is if there was no relation between the independent variables.

The two models ask different questions so they get different answers.

It's also not surprising that distractors had a large effect on a task like this.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.