basic question: surely that his graph is visually misrepresenting the data?

gene

New Member
#1
if a graph has ratio on one axis and is not logarithmic in scaling between 0-1 then surely the graph is misleading right?
that is, the ration between 0-1 are infinite, logarithmic, but after 1 to ∞ the ration is not logarithmic.
for example the distance on the graph between 0-1 is 1cm; so the ratio of 1:1 is 1cm into the graph.
the ratio of 10:1 is 9cm distant away from 1:1 on the graph, but the reverse is only ~0.7cm away from the 1:1,
this gives a skewed perspective,
this effect probably has a name
here is an example
in this paper
1610399185556.png
 

gene

New Member
#3
visually the data in the graph preceding 1:1 is skewed relative to the rest of the graph; it is bunched up before 1, therefore it is disproportionate
 

katxt

Active Member
#4
You can look at it as being linear in P for a given amount of C.
I'm not familiar with the context so I cannot say more, but how would you improve the graph so that it illustrates what you think is really the connection between Risk and Diet.
 

gene

New Member
#5
the way it is results in the gradient/steepness of the rise prior to 1:1 appearing to be steeper than after 1, as it is all bunched up between 1 and 0. surely this effect has a name
 

katxt

Active Member
#6
I find it hard to share your indignation. but perhaps it is justified if you know the context and think that the author is somehow using the graph to mislead the reader. Perhaps 0, 1, 2, 3 would have been better put as 0%, 50% 67%. 75% but that might not have got the authors point across as effectively.
A log scale squashes data unevenly and alters the slope of the graph, but perhaps you think that is also misrepresenting the data.
At any rate, I haven't come across a specific name for the effect.
 

fed2

Active Member
#7
i think it just depends how 'risk score' was derived defined. the confidence limits look symmetric on the given scale, that may provide some clue.
 

gene

New Member
#8
it was not a deliberate mislead, but surely having the ratios <1 being intrinsically logarithmic and the >1 being not logarithmic makes the data appear skewed?
 
#10
for example the distance on the graph between 0-1 is 1cm; so the ratio of 1:1 is 1cm into the graph.
the ratio of 10:1 is 9cm distant away from 1:1 on the graph, but the reverse ratio is only ~0.7cm away from the 1:1,
this gives a skewed perspective?