I want to build a scoring model to determine a client's rating (Y). Y can take 9 values (A1, A2, A3, B1, B2, B3, C1, C2, C3). I have 29 parameters, some of which are interdependent, so I'll need to take them out. Each parameter is some kind of economic indicator be it a client's wage or years of employment. In the end each is reduced to the number of scores it gives to a client, depending on a thresholds I have, and in the end all scores are summed up each with some weight.
I also have rating's values for each client estimated by humans in other department, so in the end I need to get similar ratings to those of them.
What I don't have is the data about defaults. I need to build my model on just parameter's values and the ratings that are given to me.
How would I do that? I'm fairly new to statistics and from what I read log model is used to determine whether the client would default or now - gives in the end only 2 values, while I have 9 possible values.
In the end I need to be able to get both optimal weights for each parameter and it would be good to also be able to determine the optimal thresholds (but for now, it's not as important as the weights).
What model should I use? I appreciate any help and if you can advice me on articles/other sources to read, please, do so.
EDIT: To comment about data. I have the values of parameters (financial data of the client), and I have the ratings approved by our analysts for each client. I treat their ratings as a dependent variable Y, so if I were to have the default data I would be able to make a traditional log model: but now that I have only scores and ratings I would need to make a model with scores only: