Fossils offer a unique window into the history of life on Earth, but incorporating them into algorithmic approaches for the study of macroevolutionary patterns and processes has been fraught with difficulty. Morphological data are much more difficult to collect at scale than DNA sequences, limiting the size of phylogenetic trees that can be inferred for extinct organisms, and the methods used to analyze them have long lagged behind those that are routinely applied to molecular datasets in terms of sophistication. Here, I aim to alleviate these issues by developing several novel parametric approaches to the inference of fossil phylogenies and their use in downstream analyses of macroevolutionary rates. In Chapter 1, I present a new method called Bayesian Least-Squares Supertrees (BLeSS), which takes as its input a profile of previously inferred time-calibrated trees with partially overlapping leaf sets, and returns a posterior distribution of time-calibrated supertrees on the union of the leaf sets of the individual source trees. I provide a fast and flexible implementation of BLeSS in a pre-existing software environment, and validate it using a recently introduced simulation- based pipeline. Using both synthetic and empirical datasets, I identify the conditions under which BLeSS performs well, demonstrate its ability to handle supertrees of up to 1000 leaves, and provide guidelines for its use as well as suggestions for further extensions and performance boosts. In Chapter 2 (published as Černý and Simonoff, 2023), I adopt a protocol originally developed for phylogenetic analyses of genomic data, and repurpose it for the study of morphological characters to address a controversial problem in fossil phylogenetics: the earliest divergence among dinosaurs. I show that despite substantial differences in character coding, which were thought to be of critical importance by previous studies, neither of the two examined datasets significantly favors any of the three plausible hypotheses over the other two. Moreover, each hypothesis is supported by a nearly equal number of characters, including characters with high phylogenetic signal. This suggests the datasets in question suffer from a high degree of internal conflict, and are inadequate for resolving the problem they were designed to address. In Chapter 3 (published as Černý et al., 2022), I use empirical and synthetic data to evaluate the performance of two recently introduced Bayesian methods for the inference of macroevolutionary rates from fossil phylogenies and occurrence records. I show that their ability to tease apart shifts in the rates of speciation, extinction, and fossil sampling is limited, possibly because of the paucity of data available to vertebrate paleontologists. In particular, the two approaches are unable to conclusively answer the question of whether non-avian dinosaurs entered a period of decline prior to their extinction at the K–Pg boundary. I outline best practices for informed use of model-based diversification rate estimation in extinct clades, and suggest how it can be profitably combined with more traditional paleobiological approaches.