We present a general framework of detrending methods of fluctuation analysis of which detrended fluctuation analysis (DFA) is one prominent example. Another more recently introduced method is detrending moving average (DMA). Both methods are constructed differently but are similarly able to detect long-range correlations as well as anomalous diffusion even in the presence of nonstationarities. In this article we describe their similarities in a general framework of detrending methods. We establish this framework independently of the definition of DFA or DMA but by investigating the failure of standard statistical tools applied on nonstationary time series, let these be intrinsic nonstationarities such as for Brownian pathes, or external ones due to additive trends. In particular, we investigate the sample averaged mean squared displacement of the summed time series. By modifying this estimator we introduce a general form of the so-called fluctuation function and can formulate the framework of detrending methods. A detrending method provides an estimator of the fluctuation function which obeys the following principles: The first relates the scaling behaviour of the fluctuation function to the stochastic properties of the time series. The second principles claims unbiasedness of the estimatior. This is the centerpiece of the detrending procedure and ensures that the detrending method can be applied to nonstationary time series, e.g. FBM or additive trends. Both principles are formulated and investigated in detail for DFA and DMA by using the relationship between the fluctuation function and the autocovariance function of the underlying stochastic process of the time series.