In regression analysis how to determine which is the dependent variable?

I don’t think that there’s a specific statistical test for causality as such. You probably have to look at the theory to figure it out. A regression, or statistics in general, will tell you if you have correlation between two variables, but economic theory will point to which variable is the dependant one and which the explanatory one. For example it is reasonable to assume that ice cream consumption may be influenced by the weather, but not the other way round; or if one variable is lagged, it will be the explanatory one; or as already pointer out, time will always be an explanatory variable. However, sometimes you will not be able to prove which way causality runs, or even if there is any causality, or if both variables are influenced by a third variable which has not been specifically modeled in your system.

Draw a graph. The independent variable is drawn on the x-axis and the dependent variable on the y-axis. In some cases, it is up to the experimenter to decide what the dependent and independent variables are. As x varies, how does y vary? If the scatterplot shows a pattern (increasing or decresing), you can decide which depends on which.

Some variables such as time, can only be independent, but if you consider weight and height either of them could be dependent or independent.