There are times when we need more sample to achieve a more accurate estimator. Since these two sets of sample have the information about the same population, it is necessary to treat both as a single combined data. In this paper we present the unpooled sample variance for the combined data when we just know a sample mean and variance for the each data set without the raw data. It is shown that the pooled variance $s^2_p$ is always greater than the exact variance $s^2_t$ when ${ar{x}}_n;=;{ar{y}}_m$. And the difference of means for two data, ${ar{x}}_n-{ar{y}}_m}$, is larger, the difference of $s^2_p$ and $s^2_t$ is larger.