Find the mean for the random walk. Use past data of the phenomenon under investigation. Calculate or find the average value for the first year you wish to model. If you plan to only model future values, use the average value for the most recent year.
Choose the trend for the random walk. Decide on the magnitude of drift for the random walk that moves the values in a certain direction over time. If you believe, or if past data indicates that the value will go up, choose a trend parameter that is positive. You may estimate the magnitude of the trend by looking at the average change per unit time in the previous data.
Set the seasonal factors. You need to decide on four seasonal factors -- one for each season. These seasonal factors represent the average change in value per unit time during the seasons for which they are associated. Again, you can use past data to estimate these factors.
Run the random walk model. Use statistical software to run a random walk with the factors above. Essentially, you need to set up a random walk that begins from the mean value and moves trend * random component + seasonal factor * random component each step. Create a model that states Z(n) = Z(n-1) + t*e0 + s*e1, where Z(n) is the value at time “n,” “t” is the trend term, “e0” is the first random component (chosen at random by your statistical software), “s” is the seasonal component and “e1” is the second random component (chosen at random independently of e0). Run the model for as long as you will run the time series.