Data in percentage form and log-lin, log-log model

[newstudent]

I want to estimate a combination of log-lin and log-log model, which is given in the following form:

lnY = b0 + b1lnX1 + b2lnX2 + b3X3 + b4X4 + b5X5 + u

The data on x1 and x2 variables is in normal integar form therefore i took natural log of x1 and x2, while the data on x3, x4 and x5 is in percentage form, taking logarithm will turn them negative so i use data in levels and didnt took the natural log. My questions is can I use such model as it is a "combination of log-lin and log-log model". So can I estimate this model?

[startz]

Sure, why not? Just be careful about how you interpret the coefficients.

By the way while it's true that the log of a percentage is negative, there's no harm in that.

[newstudent]

Thanks very much, but can you guide me about results interpretation. In my view, the slope coefficients (betas) of X1 and X2 are its direct elasticities, so they are easy to interpret. For X3, X4 and X5, I calculated their elasticities by multiplying the Beta (slope coefficient) with the mean value of those variables i.e. (b multiplied by mean value of the independent variable). Is it correct?