

Test Review - focus on second half of the semester but it is cumulative.

Jeff - Tufte Graphics lecture

good graphics?

Napolean graphic - good graphic

Playfair one of orig inventors of data graphics - ?wheat and wage graphic?

Another graphic - french train schedule from the cover of the book - takes the place of a table.

map of northern galactic hemispheres 1.3 mil galaxy shown

What made these graphics good?

The amount of info that can be extracted is large

You can see relationsihops between things

Time Series graphics (first two ) can see things happening over time

__________ end of good

Bad graphics - 

mining graphics - bar graphs 
	what is bad: scales are off - some things really shouldn't be compared - also base isn't really 0 - so on thing that is a loss looks like a positive number (base looks like it must really be neg 4 mill - this makes it quite misleading.


Commission payment ot travel agents: 

3D isn't necessary (oops I missed the main point - spaced out)

________________

Comparative annual cost per capita of insane in Pittsburgh City Homes and PEnn State Hospital

size of insane aslyms which represents diff between amount spent seems misleading - small diff in per cap spending is shown as big diff in building size.

------------------

Usa today - tools workers use at meetings

pic of writing utensils on yellow pad of paper - item shown doesn't rep that line of the percentages - for example a highlighter is used to rep a laptop. 

Percentages dont add up - althought they probably don't need to since someone may use more than one at meetings. 

__________________________

More example bad graphics from USA today. 

--not very data rich

-- and not accurate in relative lengths

-- discussion in class are these really so bad - you can immediately read them and get basic info - but this definitely isn't hte point Jeff wants to make

Why are these graphics so bad: bad labeling (misleading)
Unnecessary garphics
graphics done't relate to the data
 misleading scales
graphics mask relationsiops between data
not much data
expect people to draw infreences between data
establish misleaign realtionsips between data
(one more I missed)

___________________

How can you make better graphics: from Tufte

1.) Show the data when you can.(example a trade grahics by Playfair, another set of trade balances in a graphic with larger grid so even a bit easier to see the grid scale)

2.) Ink that changes as the data changes - data-ink ratio the amount of data ink over the total ink that is used to print the graphic. proportion of a graphics ink that is used to convey teh info

Looked at midterm histogram: data-ink ratio is <0.05.  - could be better... 

Looked at "reasons americans want mars missions" graphic again it has 1% data to ink ratio.

_______________

How can we make better graphics?

Above all show the data

Maximize data to ink ratio
	- within reason
	- every bit of ink on a graphic requires a reason

	- erase the non-data ink
	-erase redundant data ink

- Jeff went through process with midterm his. what can we get rid of:
	- get rid of gray background
	- get rid of horiz lines 
	- get rid of line for x  and y axises
	- just leave the numbers that actually are the value of a line 
(discussion ....)
	- reduce it even further my making skinny lines - 

Revise and edit. 

Principals applied to common forms 
	scatter plot - orig  vs modified graph - another scatter plot variant - has lines in it to make it easier to see the info.  

Conclusions: 

show data variation not design variation
avoid using nk for non-data itesm
avoid redundancey
clear and detailed labeling should be used to defeat graphical distortion. 
revise and edit.












