# Thread: 2x2x2 factorial design with a covariate (3-way ANCOVA?)

1. ## 2x2x2 factorial design with a covariate (3-way ANCOVA?)

I have a question regarding my design and how to analyze.
I have 3 independent variables all consisting of 2 levels:

1 Treatment: Interpersonal treatment versus no treatment (control group) (manipulated)
2 Facebook use: excessive fb use versus No fb use (manipulated)
3 Residential status: living at parent's home versus living on your own

My dependent variable is a score on a depression scale.
Because I measure Pre and post treatment I take pre test score as a covariate.

My researchquestion is something like:

the influence of Facebook use on depressive symptoms in first year students with different residential status who receive interpersonal treatment. This led to the following hypotheses:

Hypotheses

Hypothesis 1: Excessive Facebook use will have a larger negative effect on depressive symptoms in students living on their own then those living with their parents.

Hypothesis 2: Excessive Facebook use will have a larger negative effect on depressive symptoms compared to no Facebook use treated by interpersonal treatment in both living at home at living on their own.

(I am also not sure about these hypotheses and they may need to be restated)

So should I do a 3-way ANCOVA or planned contrast or something else?

Can some one please tell me how I should analyze this and the thought behind it. (I only need this for my research proposal assignment so I do not actually have to perform the analyses)

Thanks!

2. ## Re: 2x2x2 factorial design with a covariate (3-way ANCOVA?)

Hey camille123.

You are going to have to tell us a bit more about your data, experiment, domain knowledge and assumptions being used.

In statistics you take information and generate test statistics that have a distribution and a link to some variable [often called a parameter] that all information is connected to.

The thing you are doing is looking at links between parameters you are looking for [between information and other parameters] and trying to assess if the inference on those parameters answer the kinds of questions you are seeking to answer.

The thing with models is that the variable type will impact the technique you use to get the statistics you need [categorical data is a lot more different than when the data are along the whole real number line].

I suggest you outline these things for the readers so that we can get a bit more of an idea of what data you have and how you might use it to answer these questions.

You have talked about a specific technique along with some ideas of the data you have but there are a lot more points needed to give more specific advice.