ANOVA?

vub

New Member
#1
Hello,

First time posting. I am given a paper on prototypes and random data. Teacher told us to use T-test instead of ANOVA, yet i think ANOVA is a better testing tool. This is data given:

-24 participants
-Tested on 3 conditions
-items: 22 per condition per participant
-Measurement: reaction times

I could use t-tests to test hypothesis, but if I were to use ANOVA how would I do it?


Ho
 

vub

New Member
#2
Hello, me again.

Normally testing ANOVA is not a problem, the difficulty is having 24 participants witch each 3 conditions with 22 items.
I think I should calculate the mean reaction times per condition then calculate ANOVA which would make it a lot easier, although I think it would threaten the validity of the test... idk
 

obh

Active Member
#3
Generally, before getting to your question.
1. ANOVA with 2 samples equivalent to a two-sample test (pooled), exactly the same results, since both distributions T and F derived from the normal distribution.
2. You may use ANOVA to compare if there is a difference between several groups, but if there is a difference you won't know which is group the difference. Or you may run multiple comparisons (with correction) to see where will you have different.

So how many observations do you have per one participant? 3? 3*22?
 
Last edited:

vub

New Member
#4
Generally , before getting to your question.
1. ANOVA with 2 samples equivalent to a two-sample test (pooled), exactly the same results, since both distributions T and F derived from the normal distribution.
2. You may ANOVA to compare if there is a difference between several groups, but if there is a different you won't know which is group the difference. Or you may run multiple comparisons (with correction) to see where will you have different.

So how many observations do you have per one participant? 3? 3*22?
To answer your question: 66 observations devided into 3 conditions witch each 22 observations (items) per participant.

The hypothesis: participants (PP) will score (RT) lowest on Condition (C)1, then C2 then C3.

Second hypothesis: PP will score equal to C1 and C2 and lower on C3.

Oh yes, I totally overlooked the reason one uses ANOVA, obviously there should be a difference begween groups! It's clear now. Then I should use t-test within PP and between PP.
Hyp Alt: U1 < U2 < U3

This helped me a lot..

P.S. If you're interested, the experiment is based on prototype enhancement effect theory and examplar based theory, it is to see which theory is supported most by results or perhaps both theories come into play, but the latter I am not interested in for my specific paper. :p

Thanks for the reply!
 

obh

Active Member
#5
Hi Vub,

So you have 3 dependent (since the same participants) groups: C1, C2, C3

You may run repeated measures ANOVA, like ANOVA but for dependent groups.
Or you may run multiple comparisons of the paired t-test. (with a corrected significance level)

The test power may be a too low, depend on the effect size you want to detect.