Behavior of the sampling distribution of a robust coefficient of variation for exponential samples in the presence of outliers.

A robust form of the coefficient of variation, CVr, a unitless measure of spread constructed using the median absolute deviation and the sample median, is considered for contaminated samples from several exponential distributions. The cases studied are those involving samples from exponential distribution with parameters A~0.2, I, 5 and 10, and contaminating percentages of a maximum of 40% outliers from exponential distributions with A~0.02 and 20, and from lognormal distributions with 0=0.25 and 5. Based on the bias and root mean square error, results exhibited perseverance of the robustness property of the CVr when compared to that of the conventional coefficient of variation, CVc, across all values of A and types of contaminating distributions.

PDF ID 27740.pdf Restricted to Repository staff only Download (680kB)

Actions (login required)

View Item