# Linear regression for a survey - help needed

• Apr 11th 2008, 02:51 AM
lisaly23
Linear regression for a survey - help needed
Hi everyone,

For my master thesis, I am trying to prepare a survey on the general key success factors for virtual worlds.

My hypotheses are that user generated content, community management, etc. all those (independent) variables are key success factors. I asked questions on a 5 point lickert scale and then want to do a regression analysis in order to verify my hypotheses. However, I do not have any statement or question that measures the success of a virtual world. I cannot think of one and without it I cannot do my linear regression, right?

How can I do that though? Sorry, I am very bad at maths...

Kindest regards
• Apr 11th 2008, 06:33 AM
falcald
Interesting problem.

It would be useful if you show us this questions and this 5 lickert scale points. Just to have an idea of the information you are gathering.

On the other hand, I dare say I think your problem is not a "statistical" one. In a regression analysis, you could relate any two variables (or more) and the test could be in favour or against your hypothesis. Having this in mind, you could use a variable as a "core" of "success". For example, you may state with non-statistical fundaments that success can be asociated to quantity of visits. Well, there you have a core variable and more visits mean more success. If you relate that variable to the factors you are measuring, a regression analysis will be easy to be set.

You may do a "discriminant analysis" to understand which factors are those which determine success from a set.

I do not know if I've been clear enough. My english is still quite bad.

Federico.
• Apr 13th 2008, 11:39 PM
lisaly23

Thank you very much for your help.

Well, here is a link to my survey so you can have a look at it.

Virtual Worlds_new

If you define success as the number of active user, the problem is I first of all might not have all the data, and secondly, I do not know which virtual world to take, as I want my research to be general so that it can determine the success factors for all the virtual worlds.

• Apr 21st 2008, 06:47 PM
falcald
I've been on holidays (in Salta, Argentina). I'll review your poll from my work and then give you my not-profesional view.

I did some work related with cigarettes marks (brands) in the Argentinean smokers market. I had different marks. Smokers had to select one and answer a poll with questions like yours. We had a formula which gave points to each participant and then we ranked them with this points. For a very well-known american car manufacturer we did the same. In that cases, we compared among brands. It would be a first step to understand which factors are determinant for this virtual worlds in an ex-ante definition.

Do you have SPSS? If you have, you can do a Discriminant Analysis to understand which of the variables in your poll best explain your "rank" or a "quality" like: "3 hours per day of use", "1 hour per day", "less than 1 hour". You also can run a Cluster Analysis to understand which variables are related.

If you don't have, please send me an email and we could use another software called R ---> The R Project for Statistical Computing which is free and a powerful tool (like S-Plus) and you can use this tools.

Hope this helps.

Federico.
• May 3rd 2008, 05:04 AM
lisaly23
hi federico

I just sent you a private message.

I hope you can help me.

Bye