The interference rejection combining (IRC) receiver, which can strictly suppress intercell interference based on the minimum mean square error (MMSE) criteria, is effective in improving cell-edge user throughput. When assuming the Long Term Evolution (LTE) or LTE-Advanced downlink and open-loop transmit diversity employing the space-frequency block code (SFBC) using Alamouti coding, the IRC receiver must detect the Alamouti coded signals and suppress the interference signals using a couple of received signals in the frequency domain at the same time. To achieve this, the IRC receiver weight matrix, which consists of the channel matrix of the serving cell and the statistics of the covariance matrix, including the interference and thermal noise components, must be extended in the frequency domain, i.e., due to the effect of Alamouti coding, in addition to the spatial domain. These extended matrices can be estimated using the downlink reference signals (RSs) from the serving cell. However, some elements, including the effect of Alamouti coding in the extended covariance matrix, cannot be estimated using a practical estimation scheme that subtracts the replica symbols of the serving cell generated by the estimated channel matrix and the known RS sequence from the received RSs of the serving cell. This is because the RSs in LTE/LTE-Advanced are not transmitted using two adjacent subcarriers. This paper investigates the statistics of these unknown elements and proposes appropriate values, specifically inserting zero values, for these elements assuming the LTE/LTE-Advanced downlink. The results of simulations show that the IRC receiver using the proposed scheme, which has two receiver antenna branches, suppresses the intercell interference and improves the throughput by more than 10% compared with that for the conventional maximal ratio combining (MRC) receiver when a cell-edge environment is assumed.