What's the minimum amount of observation for simple multiple regression? I have a 16 year data for my dependent variable with 5 independent variables. is it ok to regress that? 
 A: Yes, you can regress your dependent variable on the 5 variables but how you do it may depend on the purpose of your modelling. 
If you are interested in prediction based on your model, you would probably be better off to use methods like lasso or elastic net, say (assuming your dependent variable is continuous). Which of these methods you might use would depend on whether or not collinearity among predictors is an issue for your data.  
If you are interested in estimating the effects of the predictors on the dependent variable, then you might not have any choice but to fit a linear regression model to your data (just make sure you address collinearity issues among predictors). The issue would having so many predictor variables relative to the small number of observations is that you would have low power for detecting significant effects of the predictor variables - thus expect to see large p-values for most of your predictor variables in the model summary. 
With data collected over time, the values of the dependent might be correlated over time, so that is something that might need to also be addressed.
