Questions about 'nobs' option!

I have a question about ‘nobs’. I tried ‘nobs=1’ and i can even get the estimation result, I do not understand why using 1 observation, we could even get results, only thing we could not get is the ‘smoothed shocks’, Matlab is unable to get the handle for ‘smoothed shock’ graph. And also I saw some code with quite small ‘nobs’, less than the whole observation number. But why would any one only pick a subset of obversavation to estimation??? :question:

Please, I am so eager to know this!! Very curious.