search

bsed / ala

Automatically exported from code.google.com/p/ala
0 stars 0 forks source link

Improve differentiation of highlighted points on map #587

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago 
From feedback from Catherine Noble.

It can be difficult to identify highlighted points using legend facets when 
many points are mapped. The red ring that is placed around the highlighted 
points won't work well with red points. 

This is not such an issue on scatterplots.

Can highlighted points be better differentiated from background points? 
Removing non-highlighted points may remove the context. Reducing the size of 
non-highlighted points to a minimum may be an option. With either of these 
options, a 'toggle' may be needed.

Original issue reported on code.google.com by leebel...@gmail.com on 16 Feb 2014 at 8:57

GoogleCodeExporter commented 9 years ago 
Will try automatically 'flashing' the circle highlight.

Original comment by adam.col...@gmail.com on 2 May 2014 at 12:22

GoogleCodeExporter commented 9 years ago 
What about this:

If the user selects to highlighted points, colour all points the same shade of 
black/grey and keep the highlighting circle as is (red circle). 

We could also colour the circle the same as facet colour to allow multiple 
highlights of facet value.

Original comment by moyesyside on 2 May 2014 at 7:11

GoogleCodeExporter commented 9 years ago 
I think flashing could be quite distracting and wouldn't be very useful if the 
user tries to do a screen grab.

Original comment by moyesyside on 2 May 2014 at 7:18

GoogleCodeExporter commented 9 years ago 
I've 'slept' on this and still figure flashing red circles may be the best 
solution overall. Main reason is that in the general case, you may want to 
compare highlighted points with the other displayed classes on the facet. You 
could for example highlight two or more classes that you want to compare 
geographically with other displayed classes. The only way you can do this is to 
retain class colours. I agree that in some cases, all that is needed is to 
differentiate highlights from the rest.

Original comment by leebel...@gmail.com on 5 May 2014 at 12:00

GoogleCodeExporter commented 9 years ago 
I don't think flashing is the right way to go here as it will still be 
difficult to differentiate a red flashing on top of a sea of reds, oranges, 
yellows - which is currently the case.

Original comment by moyesyside on 6 May 2014 at 12:05

GoogleCodeExporter commented 9 years ago 
Just had a chat with Lee. Lee didnt really get what I was suggesting from the 
comment above.

Here is a crude representation of what i was suggesting.
So in this the user has selected to colour by year but then to select 4 
particular years. When she selects to single out 1 or years the dots are all 
turned to grey. The colour of the highlight ties in with the colour of the 
facet value. Green = 1984 for example.

Original comment by moyesyside on 6 May 2014 at 5:09

Attachments:

GoogleCodeExporter commented 9 years ago 
Arrrrrrggghhhh.  Sorry Dave. This option looses the non-highlight 
colour/classes. I thought we had agreed to leave those alone? I thought we also 
considered just using the circle in class colour without dot.

Basically - we need to be able to highlight multiple classes while leaving the 
colour of the remaining classes so that a comparison of highlight class(es) vs 
all other classes could be achieved. If we go grey or whatever for ALL 
non-highlighted classes, we loose the ability of examining the spatial 
relationships of highlights vs EACH other class. If the user wants to, they can 
create a new layer from the highlighted class(es) and then recolour the 
original layer points to grey or whatever.

Original comment by leebel...@gmail.com on 6 May 2014 at 7:21

GoogleCodeExporter commented 9 years ago 
I dont think you are losing anything here, but making it much more usable. The 
split on all colours is there for the user if they dont choose to use the 
highlighting function.

Having 30 + colours on a map + highlights + flashing is producing a colour 
explosion which is technically impressive, but very difficult to use in a 
sensible fashion. Im sorry Lee but its bad design.

Original comment by moyesyside on 6 May 2014 at 7:54

GoogleCodeExporter commented 9 years ago 
[deleted comment]
GoogleCodeExporter commented 9 years ago 
Lets park this one until we can come up with some sensible alternatives we are 
comfortable with. We certainly have plenty more things to progress at the 
minute.

Original comment by moyesyside on 6 May 2014 at 8:39

GoogleCodeExporter commented 9 years ago 
I agree to park it for now, but to state the case as clearly as I can so it can 
be revisited when we are up to it...

 - if we have a taxa with 10 classes on one facet and we want to see what the spatial pattern is of a set of say two classes (lets say classes 1 and 2) against the other 8 classes (3-10) we need to be able to differentiate those 8 classes from the two highlighted classes. If for example, we have a spatial pattern where lets say the class 3 is occupying the NW quadrant, this will only be apparent if class 3 is identifiable.

If we truly only want highlighted classes against the remainder of 
UNDIFFERENTIATED classes, then we can do that now. You would create a new layer 
from the highlighted classes and then colour the original layer points as say 
grey or whatever and set the new layer as some other colour. You could even 
colour the new layer by their original classes if there is more than one class 
that was originally highlighted. We can fiddle with the transparency and 
priority of the two layers to achieve flexible outcomes.

I agree that it may be a sad design when there are heaps of points, but the 
current function does allow for the type of analysis just described. I'm more 
function over form in this situation as zoom/pan does allow for a lot of 
flexibility to reduce the clutter in situations where points abound. 

If we go down the grey ('background - unselected' classes) path,  achieving a 
general solution (as I put above) would also require the generation of a new 
layer of highlighted points. Those two layers would then permit the 
identification of individual background point classes (original layer) against 
highlighted classes (new layer).

Therefore, with either strategy (grey/undifferentiated  'background' classes or 
coloured 'background' classes), a general solution is available. It therefore 
comes down to what is the default. I suppose the simpler 
'grey/undifferentiated' state makes sense as the more general solution would be 
suited to more advanced users.

I do like the idea of colouring the highlight circles as per their class 
colour. As we both agreed, maybe slightly thicker so the colour of the class is 
obvious. Using the circles alone as I suggested, would seem to me to simplify 
the potential complexity of the display. This would work with either 
strategy/default.

Either way, it seems like I write more help.

Original comment by leebel...@gmail.com on 6 May 2014 at 9:23