Transient stability assessment involves the determination of whether all generators in the power system would maintain synchronous operation when it is subjected to a large disturbance. Assessing the transient stability status of a power system in an accurate and timely manner is of high importance in order to plan and initiate relevant corrective control actions where required. The wide-area monitoring systems enable to acquire synchronized measurements of important system variables from geographically dispersed locations via the phasor measurement units (PMUs), which can be used for measurement-based transient stability assessment. In this work, a stacked ensemble model that uses post-contingency wide-area synchrophasor measurements of bus voltage magnitude, voltage angle, frequency and rate of change of frequency is proposed to classify the post-contingency transient stability status of a power system to be unstable or stable. The model includes four individual long short-term memory network classifiers as base-learners and a LightGBM classifier as a meta-learner. An important differentiation of this work is that the proposed method does not assume that each bus is equipped with a PMU. Furthermore, robustness of the proposed model is analyzed in the presence of measurement noise and against the scenario of missing measurements. The performance of the proposed ensemble model is investigated in two generic test systems, the Nordic and the 127-bus WSCC test systems, as well as the practical Turkish power system.