How to perform residual analysis for weighted linear regression? How do we perform residual analysis (verifying homoskedasticity, normality and independence of errors)  for weighted linear regression? By weighted, I mean each row in the dataset has weight assigned to it.
 A: We can perform it in almost the same way as for unweighted regression, except that, since regression variances are inversely proportional to weights, standardized residuals (for example) must be multiplied by $\sqrt{w_i}$, giving what's sometimes called weighted standardized residuals. 
Similar (but slightly more complicated) effects come into calculating the equivalent studentized residuals, yielding weighted studentized residuals.
Statistics packages will calculate weighted standardized (/studentized) residuals for you.
So you just check normality, or do plots of standardized residuals vs fitted or whatever other diagnostics you like.
However, you can't verify homoskedasticity of the conditional observations because the data should actually be heteroskedastic. You can verify that the weighted residuals are homoskedastic, though, which implies that the conditional observations have the variance implied by the weights.
Typically, you can't really check for general dependence (weights or no weights), unless the data have special structure that suggests a particular form of dependence to check for (e.g. if the data are are ordered - in time or space, say - you can check for serial dependence).
