Reliable in situ thermal characterisation allows to study the actual thermal performance of building components rather than the theoretical performance calculated from thermal properties of the constituent material layers. The most generally accepted method for in situ thermal characterisation is the average method as described in ISO 9869. However, due to steady-state assumptions, the method's applicability can require long measurement periods and is often seasonally bounded. A correction for storage effects might shorten the required measurement time spans for the average method, but will not eliminate the seasonally bounded limitations. More advanced dynamic data analysis methods, such as regression modelling, ARX-modelling or stochastic grey-box modelling, can be used to overcome these difficulties. In this paper, a comparison between several semi-stationary and dynamic data analysis methods typically used for the thermal characterisation of building components from on-site measurements is made. Thereby, special attention is given to the reliability of the methods thermal resistance estimates when confronted with data sets of limited measurement time spans and different seasonal boundary conditions. First, the methods' performances are assessed for simulated measurements of a south-facing insulated cavity wall in a moderate European climate. Subsequently, the performances are examined for actual measurement data of a similar test wall. (C) 2016 Elsevier B.V. All rights reserved.