# Selecting the right type of regression analysis method.

#### About_ to_ graduate

##### New Member
Hi,

My knowledge about statistics is elementary and I would really appreciate some help or suggestions in solving my current problem. I am writing my dissertation paper and I am struggeling with which type of regression analysis I should use. I collected the data using a questionnaire with the structure shown bellow. The objective is finding out the relation of some factors with the # of ideas implemented by the employees. What you recommend me? By the way I am using SPSS for the data analysis.

Scenario:

Dependent variable:
Question: Qty of ideas implemented: (Refers to the number of ideas that the employees have implemented, e.g an employee can have 20 ideas implemented and another one can have 2)

Creativity (factor 1)
Q1: (strongly disagree=1, disagree=2, neutral=3, agree=4, strongly agree= 5)
Q2: (strongly disagree=1, disagree=2, neutral=3, agree=4, strongly agree= 5)

Risk taking (factor 2)
Q1: (strongly disagree=1, disagree=2, neutral=3, agree=4, strongly agree= 5)
Q2: (strongly disagree=1, disagree=2, neutral=3, agree=4, strongly agree= 5)
Q3: (strongly disagree=1, disagree=2, neutral=3, agree=4, strongly agree= 5)
Q4: (strongly disagree=1, disagree=2, neutral=3, agree=4, strongly agree= 5)

Openness (factor 3)
Q1: (strongly disagree=1, disagree=2, neutral=3, agree=4, strongly agree= 5)
Q2: (strongly disagree=1, disagree=2, neutral=3, agree=4, strongly agree= 5)
Q3: (strongly disagree=1, disagree=2, neutral=3, agree=4, strongly agree= 5)

and so on...

I am planning to combine all the question under each factor.

I highly appreciate your help! Thanks in advance #### j58

##### Active Member
By "combining" the questions under each factor, I assume that you will be adding them together or averaging them, which means you're treating them as quantitiative. Your dependent variable is count data, so Poisson (or negative binomial) regression might be reasonable. And I wouldn't rule out linear regression. I would probably try both and use whichever fit the data better.

#### About_ to_ graduate

##### New Member
Hi J58 thanks for your help!
I ran the Kolmogorov-Smirnov (KS) test for the dependent variable to verify if this data follows the poisson distribution and I got a p-value of 0.005, therefore this data does not follow such distribution. what do you recommend me to do in this case?

#### j58

##### Active Member
The Poisson regression model assumes that the dependent variable is Poisson distributed only under the null hypothesis. So, if your regressors affect the dependent variable, we would not expect the dependent variable to be Poission distributed. That said, count data is usually either under- or overdispered, and therefore it is usually a good idea to include a dispersion parameter in the model. This can be accomplished by using negative binomial regression instead of Poission.

#### About_ to_ graduate

##### New Member
Hi j58, your clarifications has been really helpful so far, but I haven't been able to find the answer yet. If it is possible and is not too much for you, Can you check out this data and tell me how can I untangle this mess? Thank you very much and sorry for the inconveniences ! The first column of the data is the dependent variable. The rest are the independent ones.

#### Attachments

• 19.9 KB Views: 4

#### j58

##### Active Member
ATG, I'm happy to give general advice online, but I have to stop short of working with your data. If you're interested in hiring me for statistical consulting, send me a private message.