In surveys, the voluntary involvement is particularly vulnerable to non-response, especially when a critical topic is involved. Response rates for COVID-19 were also low throughout the world since people feared for their lives based on viral myths among them. This low response rate can produce bias in the population estimates which can affect precision of the statistical analysis. To overcome this issue, we suggest a novel sub-sampling technique under ranked set sampling (RSS), which we name the substitution method. The estimator of a population mean is developed under the substitution method and expressions for its bias and variance are derived. A theoretical efficiency comparison of the substitution method with respect to existing sub-sampling methods is provided, and a simulation study is performed to support the theoretical findings. Additionally, the difference estimator is also proposed based on the substitution method when non-response occurs only on the study variable. It is also shown that the difference estimator estimate population mean more precisely under the substitution method when compared with existing sub-sampling methods under non-response. For real-life application, COVID-19 data of Pakistan is also considered for simulation study.