The study suggests a class of regression type estimators for estimating the population mean when there exist either a positive or negative correlation between a study and auxiliary variable. The auxiliary attributes have been used in the form of the standard deviation, coefficient of kurtosis, mean deviation, quartile deviation, median and 10th decile. The bias and mean squared error has studied up to the first order of approximation. A comparison is made both theoretically and numerically with some existing regression estimators. It is observed that the suggested class of estimators provides more reliable and efficient results than others.