Skip to Main Content
The well-known sample variance estimator utilizes N samples from a random process to first estimate the process mean. The estimator then uses the same N samples to estimate variance from this mean. Process variance could also be estimated by first using less than N samples to estimate the mean, followed by using all N samples to estimate variance. Two-scale estimators of this type, both causal and noncausal, are defined. Statistics for these estimators are derived, which are valid for samples from any statistical distribution. These statistics are used to improve analysis of a previously reported device called the double detector. In microwave radiometry, the double detector senses the presence of deterministic signals, often called radio-frequency interference, that corrupt the usual measurement consisting only of Planck radiation.