# 2IVs, multiple-levels; 1DV=frequency

#### Lucas Glover

##### New Member
EDIT: (in title I meant 2 IVs, not 2DVs!) - I fixed it for you - Dason

Hi,
I have a question: what test would be appropriate?

IV = genotype, 2 levels (wild type, mutated)
IV = behaviors, 6 levels (climbing, etc..)
DV = frequency (how often does each behavior occur)

Last edited:

#### noetsi

##### No cake for spunky
Re: 2DVs,multiple-levels; 1DV=frequency

If you have at least 10 different levels of the dv I would just do OLS. It's easy to do and interpret and commonly known. If you have special needs for the analysis, or you don't meet the assumptions of the method that would not work of course.

#### Lucas Glover

##### New Member
Re: 2DVs,multiple-levels; 1DV=frequency

I have 8 levels of my IV, behavior
My DV is frequency ((how many times this behavior occurred in each type of mouse (wild type vs. mutated).

In your post, did you mean "...at least 10 different levels of the [iv] ..."?

#### noetsi

##### No cake for spunky
Re: 2DVs,multiple-levels; 1DV=frequency

How many levels of your IV you have does not really matter in choice of method. Only the number of levels of the DV.

#### Lucas Glover

##### New Member
Re: 2DVs,multiple-levels; 1DV=frequency

Well, if levels of DV, just one, I do suppose, as I am just counting frequency

#### noetsi

##### No cake for spunky
Re: 2DVs,multiple-levels; 1DV=frequency

No you will have one level for every possible distinct value frequency could take. For example if there were a 100 distinct levels (1,2,3,4....) frequency could take on in your data you would have a 100 levels.

#### Lucas Glover

##### New Member
Do you mean the count of the frequency for each level of the IV (behavior) or all counts all together for all levels of IV? Did you look at the graph I posted on a website?

#### Dason

I'm not sure what noetsi was trying to get at. But can I ask you to describe your data a little more? Specifically do you only have one observation for each combination of genotype and behavior? Also - what are the research questions you're trying to answer here?

#### noetsi

##### No cake for spunky
I did not understand the graph in honesty. I meant all possible values that the DV can take on in your data. Which should be all counts for all the levels of the IV (you can just count the distinct different levels of the DV in your data and you will know how many levels there are). A level and a value of the DV are the same thing (commonly you use levels when there are a few and don't use this term when you have many possible levels but that is meaningless. Temperature is a variable with many many levels).

I would be surprised if this is not interval in nature.

#### Dason

Why are you worrying about the levels of the dependent variable exactly?

#### Lucas Glover

##### New Member
I see
for both:

the graph I uploaded represents (on y-axis) the frequency # the mouse ( either tk and wt mouse breed) displayed that particular behavior (x-axis).

Situation in which I collected the data: the mouse was placed in a box and behavior was observed for 15 minutes. Every time they did, say, a "wall climb" (one of the behavior levels on the x-axis), I put a 'tick' by their name. Every time they displayed a "rearing" behavior, I put a 'tick' by their name, etc. You can see that, none of the tk mice displayed 'rearing' behavior, so they had a frquency of 0, whereas the wt only showed 2 'rearing' behavior.

The data bars represent the mean from all the wt (left bars) or tk mice (rght bars).
I had n=7 tk mice' n=4 wt mice.

EDIT: what I want to see/test is whether tk or wt mice differed significantly in any of those behaviors? not only across the behaviors but more specifically the genotype ((wt or tk). Essentially I really want to look at, for example, whether wt mice differed sig. from tk mice in, say, "stretch attend" behavior

EDIT2 Normally, if my dependent variable was a continuous variable (like "time") I would run a two-way ANOVA. This is a non-parametric situation and wasn't schooled greatly in non-parametric tests

Last edited:

#### noetsi

##### No cake for spunky
Why are you worrying about the levels of the dependent variable exactly?
to know if you can use OLS type methods (that is is the dv interval so you can apply parametrics).

#### noetsi

##### No cake for spunky
I see
for both:

the graph I uploaded represents (on y-axis) the frequency # the mouse ( either tk and wt mouse breed) displayed that particular behavior (x-axis).

Situation in which I collected the data: the mouse was placed in a box and behavior was observed for 15 minutes. Every time they did, say, a "wall climb" (one of the behavior levels on the x-axis), I put a 'tick' by their name. Every time they displayed a "rearing" behavior, I put a 'tick' by their name, etc. You can see that, none of the tk mice displayed 'rearing' behavior, so they had a frquency of 0, whereas the wt only showed 2 'rearing' behavior.

The data bars represent the mean from all the wt (left bars) or tk mice (rght bars).
I had n=7 tk mice' n=4 wt mice.

EDIT: what I want to see/test is whether tk or wt mice differed significantly in any of those behaviors? not only across the behaviors but more specifically the genotype ((wt or tk). Essentially I really want to look at, for example, whether wt mice differed sig. from tk mice in, say, "stretch attend" behavior
To me, dason is a better judge, this looks like interval data. You can only (legitimately this is often violated) create means with interval data. One think I am not sure of is that you are doing counts which are discrete and I dont know how exactly that impacts whether the data can be used in methods like regression.

#### Lucas Glover

##### New Member
I don't know the precise definition oof "counts", as it may be used differently in the statistics realm than I am maybe using it... All the dependent variable is: if the mouse displayed the behavior, it got a 'check' in a box.. If it performed that behavior 20 times, it got 20 'checks' or 'ticks' or whatever for that trial. Frequency = how many times it was displayed in the 15 min interval. The graph bars just represents the mean frequency from wt and tk mice.

EDIT: the nature of the DV is liken to sampling 300 people and asking how many people like "vanilla" or "chocolate" or "strawberry" etc.. that type of frequency at which it occurs (obviously this isn't my test, but only the nature of the DV, trying to give you what I mean from when I say "frequency")

#### Dason

Edit:

What exactly is your research question? I don't think you ever answered that.

#### Lucas Glover

##### New Member
My research question: do wt or tk mice display different amounts of each behavior. tk mice are impaired in a brain region, while wt mice are just normal mice

EDIT: so, you may have surmised, does damage in this specific brain region (in the tk mice) impair the normal amount of behavior shown (using wt mice as the controls).

EDIT2: and regarding the interval DV question: there IS an absolute zero, such that they just did not display the behavior at all. so would it be considered Ratio data? Since it is ratio DV, is two-way ANOVA okay?

Last edited: