Model for: 1 categorical and 1 continuous variable as predictors, continuous variable to be predicted I need to predict the value of a continuous variable based on one categorical (yes/no) and one continuous variable. What would be the ideal model here?
Should I do a linear regression separately for the two levels of the categorical variable, or is there a better way to put everything in one model?
 A: This is a difficult question, because there is a possibility of nonlinear behavior and different nonlinear behavior in each group. Nonetheless, you seem  like you would be happy to do a linear regression for each group separately, so it is reasonable to think that you want to do a linear model.
I see two options. The first forced both groups to have the same slope.
$$
y = \beta_0 +\beta_{1}x_{1} + \beta_2x_2\\
x_1\in\{0,1\}\\ x_2\text{ is continuous}
$$
In R software, this would be something like lm(y ~ categorical_variable + continuous_variable).
This model will give you parallel lines. If you want to allow the two lines not to be parallel (as two separate regressions would give you), then you would include an "interaction term" between $x_1$ and $x_2$.
$$
y = \beta_0 +\beta_{1}x_{1} + \beta_2x_2+\beta_3x_1x_2\\
x_1\in\{0,1\}\\ x_2\text{ is continuous}
$$
In this setting, $\beta_3$ concerns how different the two slopes are. In R software, you would do something like lm(y ~ categorical_variable + continuous_variable + categorical_variable*continuous_variable).
