In this paper, we present a cluster parameter based analysis of an outdoor-to-indoor (O2I) Multiple-Input-Multiple-Output (MIMO) measurement campaign carried out at 3.5 GHz in an office building. The measurement was performed using a wideband multi-antenna radio channel sounder. Besides azimuth angle of arrival (AoA), azimuth angle of departure (AoD) and delay domains included in traditional 2 dimensional (2D) MIMO channels, 3 dimensional (3D) MIMO channels also consider elevation angle of arrival (EoA), elevation angle of departure (EoD) when describing clusters. By introducing the elevation domain, we observe that multipath components (MPCs) within one cluster in 2D MIMO channels can be divided into several clusters in 3D MIMO channels and each cluster has MPCs with smaller power weighed multipath component distance (MCD). We also analyze the effect of height difference between transmitting and receiving ends on the clustering performance. There is a tendency that clusters show smaller parameter values when the height difference is smaller. After that, we describe all the cluster parameters by a set of probability density functions (pdfs) and give typical values in this scenario.