Issues with banking functions

[jrnold]

Issues raised by a user in an email with the functions in banking:

(1) you take the atan() of raw slopes, but in Cleveland (1993) the weighted average absolute orientation criterion is with normalised slopes, that is they are scaled by their range

(2) you calculate the length of line segments as independent of the aspect ratio, but Cleveland (1993) clealry shows the weight as a function of the aspect ratio

(3) why do you optimize the square of the mean (^2) ?

and

explain why you omitted to square the LOR and GOR criteria in bank_slopes_gor and bank_slopes_lor, as shown in Heer and Agrawala (2006)

[jrnold]

1. I take atan(slopes / alpha) which effectively normalizes them.
2. That was a mistake - I forgot to normalize them.
3. That was to take a squared distance between the mean and 45 degrees. I could have used a root finder instead.
4. It was a mistake to omit those.

After working through the code, I had too many issues getting the various methods involving optimization to find their maxima. I'm removing the methods for lor, gor, and ao. If they are specified, the function will use ms instead with a warning.