# ANOVA for different metrics

I did a test on a group of people to measure their performance. Now I have specific targets for each task and the actuals. The metrics are Time spent on the task, number of clicks and number of scrolls. I was thinking of using ANOVA to compare those but I am wondering if I can since I am using a different kind of metrics and formats (time is in minutes, clicks and scroll are n number of times). My table looks like this:

        Target  Website A   Website B   Website C

Time     00:34      01:30       01:41   01:51

Clicks     12        16       13.5      11.87

Scroll      1        2.5       6.9        3


Is it possible to use ANOVA for all of them together or I would have to do a comparison only for time, another one for clicks and another one for scrolls?

• ANOVA has a dependent variable and one or more independent variables. That doesn't seem like your case at all. Sep 3 '17 at 14:30
• Thanks Peter! What would you recommend me to do then? MANOVA? or just do separate anova for time against target, click against target, etc Thanks again Sep 3 '17 at 14:33
• From the question is not clear what you are trying to get. To select a tool, you need a research question you want to answer with that tool.
– Pere
Sep 3 '17 at 14:37
• Thanks. My question is if there is a significant difference between the target time and the time it took to make the tasks. Ho would be there is no difference and alternative that there is. Sep 3 '17 at 14:39

I don't think ANOVA or any associated method is right here. You say you want to compare actual values of some metric to the target values. So, I'd start by looking at various plots of the differences. Then you could calculate descriptive statistics on these differences.

If you want to do some statistical test, you have to set up some null and alternative hypotheses. I don't see any sign of these in your question and I don't think there really needs to be any.

• Thanks again Peter! I just want to understand two things: 1° is there a significant difference in the group's performance based on time. Ho = there is no significant difference between the groups 2° is there a significant difference between the targets and the actuals . . Ho = there is no significant difference between the target and actuals Sep 3 '17 at 14:46

I'm trying to ask to "My question is if there is a significant difference between the target time and the time it took to make the tasks. Ho would be there is no difference and alternative that there is", as stated in comments.

The obvious short answer is that you can do a t-test, with $H_0$ that mean is 00:34 and $H_1$ that mean is different that 0:34 (or larger than 0:34, depending on what you want).

However, there are several caveats.

In first place, your sample is tiny. With a 3 cases sample, you need to assume normality of your variable (which isn't unreasonable but it's a bit daring if you don't have more information) and power will be small. However, since measured times seem very different from target, you might actually reject the null hypothesis.

In second place, if you want to perform t-test on all variables (time, clicks and scroll), beware of multiple comparisons problem.