What kind of regression model would best suit this scenario?

Question

I'm new to statistics so this may seem an obvious question for you guys.
I have created a survey to collate peoples opinions on how U.S anti-money laundering regulation has impacted their financial privacy.
The survey is completely anonymous if you are interested and found here:
https://forms.gle/9Fwm1nH9E8P4ioPR9
So I am trying to model financial privacy loss (dependent variable) against U.S anti-money laundering regulation such as the Patriot Act / Bank Secrecy Act ...(the independent variables). In total I have 4 independent variables or 4 pieces of regulation.
In my model the independent variables can impact the dependent variable in 3 ways:  Weak, Moderate, Strong.
From what I understand this scenario is trichotomous and not binary, and also discrete not continuous as there is no in-between)
For example: The Bank Secrecy Act has had a strong impact on financial privacy loss.
So my question is: What kind of regression model would best suit this scenario?
Thanks for your time.

Akylas Stratigakos · Answer

One example would be this to create a vector named 'Bank Secrecy Act' and use it as an independent variable. This variable is categorical taking values in the set ${weak, strong, moderate}$. You can use one-hot encoding to create dummy variables, which can then be used as an input in a regression model. Since you have 3 states you need 2 dummy variables. One example would be:

$[0,1]$ indicates $weak$
$[1,0]$ indicates $strong$

The absence of dummies, $[0,0]$, is the base case. In this example this correspond to $moderate$ effect. Note that most packages create these implicitly, so you only need to create the input vector prior to fitting the model.

What kind of regression model would best suit this scenario?

One Answer

Add your own answers!

Ask a Question