Well I am working with a large scale national level survey data on Household Consumption , (by NSSO India) which after cleaning gives rise to almost 1 lakh households. I am attempting OLS, MNLogistic and Probit Regression Models on this dataset. Now my question is: DO I NEED to use the stata command -svyset- and declare it as a survey data and then use -svy- before every model ? Is it mandatory for such datasets? What if I do not do that and treat it as any other dataset? Will my results vary significantly?

Also, I do not clearly understand how to devise FPC etc.even if I want to declare it as survey data. My data has FSU(First stage unit) serial numbers. Is it the same as PSU ? But then they divided the survey data mainly by means of sector: rural and urban. How to go about it? I am confused and doing my PhD with not much prior exposure to large datasets.
Also, my dataset's "weight" is computed using an estimation procedure mentioned in the readme file of the survey data, where it is something like gen weight = 1/MLT if Nss=nsc where MLT, NSS and NSC are variables.So when I do descriptive statistics using the variables should I use this "weight" as Pweight or Aweight or Iweight and why?