The thesis is devoted to evaluating dielectric constant measurement models for highfrequency low-loss dielectric materials and to constructing measurement models whose application reduces measurement uncertainty when conventional measurement methods cannot accomplish it. The application of high-frequency dielectrics in modern microwave equipment is so extensive that not only universities and research centers but also companies that manufacture measuring equipment and certified test laboratories have been conducting research in this field for many years. Although universal methods and models are available, little attention has been paid to the evaluation of the suitability of the measurement model, particularly in cases where a non-destructive method is used and the model parameters must be fixed (expected value of sample dielectric constant, dimensions, shape, frequency, etc.). The use of an inappropriate model without an a priori evaluation of its suitability may lead to such a high uncertainty that the measurement results would not be usable in practice. Some models may also fail to give meaningful results for samples that, for various reasons, may not be altered. In the thesis, it is demonstrated and justified through calculations that in the case when the measurements are performed using a model with a fixed set of parameters, and the model evaluation shows that it is not suitable, it is possible to reduce the measurement uncertainty via extending this model by adding to it one or more additional dielectric objects. Another approach is based on altering the dimensions of the sample, leading to its destruction, which is not always acceptable and possible. The thesis shows the importance of ascertaining if a selected measurement model is capable of providing sufficiently high dielectric constant accuracy for a given set of model parameters. To that end, a simple and effective measurement model evaluation method is proposed - the model sensitivity analysis method. The measurement models investigated in the present thesis employ the reflection method to retrieve the dielectric constant and involve flat dielectric samples fitting tightly in the crosssection of a rectangular waveguide, flat dielectric samples located in free-space, and cylindrical dielectric full-height H-plane samples in a rectangular waveguide. The measurement uncertainty is estimated using two international standard recommended methods, namely, the Error Propagation Method and Monte Carlo Method. The latter is highly computationally demanding while providing reliable uncertainty estimation of nonlinear models. In the thesis, an efficient computational method developed by the author is used, which for models involving dielectric cylindrical rods in a waveguide, ensures highly accurate results while being considerably faster than the existing general-purpose methods, thereby allowing one to perform the Monte Carlo-based measurement uncertainty estimation for the rod-based measurement models in a reasonable time-frame, which would be impossible if general-purpose methods were employed due to prohibitively large computational burden.