Cylindrical data is a form of bivariate directional data, where one compone
nt is measured on an angular scale and the other is linear. This data can a
rise in some manufacturing applications with circular parts, such as the au
tomotive industry with wheel, brake and engine components. Also, observatio
nal data of this variety exists from meteorological and biological applicat
ions. One common application is wind direction coupled with some other char
acteristic of weather, such as temperature, wind velocity or ozone concentr
ation. Several models, which all assume good spread of the angular componen
t around the range of values, have been proposed to examine this type of da
ta. Possible limitations of the models when the angular component of the da
ta falls in a narrow range of values are considered, and an alternate model
for this case is proposed. Strategies for choosing between competing model
s based on the spread of the angular data and model fit are discussed.