Linear Regression for Forecasting: what are the risks?

Question

I have a dataset containing financial data of multiple firms on 7 to 10 years. (yearly data). For each firm/variable i want to predict its value in the next two years.

I don't have enough data (7 to 10 observations for each variable) so that's not enough for time series analysis (can't capt trend or seasonality). I think a linear regression might do, but i'm not aware of the risks? what kind of error might result on working with regression?

This is a similar question yet it doesn't really answer my question

Answer 1

you can use tslm function from RPackage forecast. Of course you should look at the p-value of the model,but not only that. your R² and R² adjusted should be close to 1. well it depends on you data, some times R² that is > 0.5 is enough. you should also test the autocorrelation of the residuals, which is possible whith the Afc function, and the normal distribution of residuals using q-q plot (there are other ways.

If all the tests are good, then the model is valid and you can apply the linear regression to forecast the data over time. The only risk that, the shorter your time series is, the shorter period you'll be able to predict with acceptable error.