"Cherry picking" is the practice of collecting data which supports your point while systematically ignoring all contradictory data (cf. Damning the Alternatives). This is a blatant violation of the scientific method and among scientists is generally considered to be unethical.
A political example occurred in the 1990s, when the "black church burning epidemic" captured headlines:
- The CDR had systematically ignored fires set by blacks and those that occurred in the early part of the decade; it had also labeled some fires as arson that clearly were not — all in an apparent effort to make black church torchings appear to be an escalating phenomenon.
Even science journals have been fooled:
- Consider a report by three environmentalist authors back in 1998 in Journal of the American Medical Association (JAMA), analyzing male-female birth ratios between 1970 and 1990. The authors found male births declining, and predictably blamed man-made chemicals. Yet public data going back to 1940 showed gender ratios are always changing, for no obvious reason. Years that disproved their thesis were simply sliced out.
- "... over the past 35 years ... A large increase was seen in the number and proportion of hurricanes reaching categories 4 and 5." 
Here is the rebuttal:
- "... not a trend. Rather, it is likely a part of the large interdecadal variations in the number of intense typhoons related to similar temporal fluctuations in the atmospheric environment." 
- "Cherry Picking in Science" - at Wikipedia