I am having some difficulty trying to get a consensus on how to draw a 
cumulative frequency curve (ogive) using grouped data. The two items of 
contention are:

1. Points are plotted cf to upper class boundary, the latter being half way 
between the upper class limit of one class and the lower class limit of the 
next class? Oxford GNVQ Core Skills "Application of Number", Gaulter and 
Buchanan, OUP 1994 confuses ucl with ucb, I think. However, on the cf axis, 
they say the median student in a hundred is the 50th, but "The strict 
definition would of course be the 50.5th student, but this accuracy cannot 
be obtained from a graph, and it is unnecessary at this stage". (??? - this 
confuses me even more - we award marks for candidates in the CSEC here in 
the Caribbean- GCSE equiv. - for plotting to 0.5 of a 2mm square on the 
x-axis, so why not on the y-axis, or are basic numeracy adults in the UK not 
up to that sort of work?).

2. When drawing the graph, are the points joined with short straight lines, 
or should we use french curves? The Excel spreadsheet seems to use short 
straight lines. School text books say a curve, and so does the cited OUP 
book above. Searching websites produces even more confusion, with a variety 
of options. I have still not gained a proper understanding of the anomaly 
explained in my earlier post (3/9/06), where some poor soul at the school 
sports day does not get a prize, because the ogive was drawn with a curve.

I have never had a great respect for statisticians, statistics seems to 
undermine the absolutism of mathematics (the most 'exact' of the sciences, 
they say). This sort of inconsistency only re-enforces by opinion. At higher 
levels, is there a continuous functional relationship between the 'x' and 
'y' values on an ogive, whereby a curve would be justified? If so, what is 
it? Can it be written here in ng text?  Kind regards, BarryAC.