Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It only takes a minute to sign up.
Sign up to join this community
I have some data about music styles played at festivals. It looks like this:
Based on this (and other) data, I'd like to make a model to explain the price of a ticket for each festival. However I'm not sure how to handle this data in SPSS. Should I treat each style as a variable? And if I did, the styles aren't normal distributed. I can't take the natural logaritm to transform the data since there are zeros in every style.
How should I proceed?
Based on this description I might consider a model of the form (where $\text{group}_i$ refers to a music category):
$$\text{Price} = \beta_0 + \beta_1(\text{Total Number of Artists}) + \Sigma \lambda_i(\text{% of group}_i) + \Sigma \eta_i(\text{% of group}_i \cdot \text{Total Number of Artists}) + e$$
Where $\Sigma \lambda_i(\text{% of group}_i)$ represents a set of variables that is the percentage of each music style and $(\text{% of group}_i \cdot \text{Total Number of Artists})$ is the interaction between these values. Note that one of the groups needs to be ommitted in each of these steps to avoid perfect collinearity.
The fact that the independent variables are not normally distributed is not an issue (only the error term matters in OLS). This specification allows you to test the marginal impacts of adding additional acts or changing the composition of the current set of acts.
As always it might be the case that any of these explanatory variables have a non-linear relationship to Price, and each can be checked either in the marginal relationships or in partial residual plots to determine if non-linearity might exist.
