I have recently had to analyse some data that was not coded correctly and although I am pretty sure its ok now I was wondering if there are any "rules" when deciding what to reverse code as well as which numbers to to assign to the extremes of the response anchors used. Also more importantly if I use a 5-point likert scale (strongly agree through strongly disagree) should all my constructs use the same response anchors, and should I consistently use 1 or 5 to denote strongly agree / strongly disagree. In other words is there any way to choose whether strongly agree is a 1 or a 5 when coded? This question probably sounds a bit simplistic, but I have found that coding incorrectly can drastically change the results when creating PLS models for example.