# Conditions for groups in independent samples test

I wish to conduct an experiment where I have a number of different groups, let's say four groups of X people each. Each person in each group partakes in a similar experiment, with each group testing a different value of the variable I want to test.

For example: Everyone plays a game of Pacman. For each group, the ghosts have different colors. I want to research if these colors affect the player's performance (score).

Now my question is first, is this possible in any way? And if it is, what conditions do I need to fulfill? I assume the groups need to be similar, so I can't have one group consisting of only students, and one group consisting of only 90-year-olds, but what would the exact criteria be?

I could also let everyone test each of the four settings, but I'm worried previous tests will affect the results of latter ones (for example, people get better at Pacman and perform much better the fourth time). Besides that, each experiment would take four times longer.

So is it possible to compare different groups with a slightly different scenario each? If so, how? If not, what would be a better option?

• You ask two questions: "is this possible" and "is this possible." But what, exactly, are the referents of the two "thises"? What are you actually asking? – whuber Jan 3 '16 at 13:57
• @whuber My main question is: Is it possible (scientifically justified) to compare results between 4 different groups, when I don't want to compare the groups themselves, but the variable (which is different for each group). And if this is possible, are there conditions the groups need to fulfill? I edited the question to hopefully be more clear. – Wouter Florijn Jan 3 '16 at 14:15

## 1 Answer

It looks like you are looking for an one way fixed effect anova. I took some time to look for a reasonable explanation and hence making it an answer (anova is a bigger terrain and the references on this site are many), the following link gives a good summary: https://en.wikipedia.org/wiki/One-way_analysis_of_variance

• Thanks for your answer. If I understand correctly, the section "Assumptions" contains the most relevant information for my question. Is this correct? – Wouter Florijn Jan 5 '16 at 17:40
• Yes, that would be your main concerns. – spdrnl Jan 6 '16 at 19:55