Summary	  Data	  coding	  ,	  analysis,	  archiving,	  and	     sharing	  for	  open	  collabora9on	                  Rich...
1.	  	  What	  is	  your	  hypothesis?	  •  9/11	  occurred	  because	  the	  intelligence	     community	  suffered	  from...
2.	  	  Observa9ons	  are	  DVs	  •  Are	  the	  paGerns	  you	  “see”	  the	  ones	  that	  are	     “relevant”	  or	  ca...
3.	  	  How	  expand	  hypothesis	  space?	  •  If	  large/standard	  datasets,	  then	  evalua9on	     becomes	  stagnant...
4.	  	  When	  do	  you	  give	  up?	  •  Reliance	  on	  visual	  paGern	  recogni9on	  by	     human	  coder	  may	  not...
5.	  	  Rules	  of	  sharing	  •  When	  does	  “your”	  data	  become	  accessible	  by:	      –  Your	  collaborators	  ...
Upcoming SlideShare
Loading in …5
×

Aslin.discussion

733 views

Published on

Published in: Technology, Education
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
733
On SlideShare
0
From Embeds
0
Number of Embeds
329
Actions
Shares
0
Downloads
4
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Aslin.discussion

  1. 1. Summary  Data  coding  ,  analysis,  archiving,  and   sharing  for  open  collabora9on   Richard  Aslin   University  of  Rochester  
  2. 2. 1.    What  is  your  hypothesis?  •  9/11  occurred  because  the  intelligence   community  suffered  from  a  “failure  of   imagina9on”   –  BoGom-­‐up  data  mining  (“connec9ng  the  dots”)   –  Top-­‐down  predic9ons  (“what  are  vulnerabili9es??”)  •  Clearly,  you  need  both  •  Must  apply  approaches  itera9vely  and  repeatedly  
  3. 3. 2.    Observa9ons  are  DVs  •  Are  the  paGerns  you  “see”  the  ones  that  are   “relevant”  or  causal?    •  Problem  of  data  sparsity  and  false  correla9ons  •  Hypothesis  tes9ng  requires  an  experiment   (manipula9ng  an  IV)  •  Tension  between  “ecology”  and  “control  of   variables”  (sociology  of  preferred  methods)  
  4. 4. 3.    How  expand  hypothesis  space?  •  If  large/standard  datasets,  then  evalua9on   becomes  stagnant  (only  evaluated  with  that   dataset)  •  If  evalua9on  only  uses  standard  (sta9s9cal)   tools,  same  problem  of  stagna9on  •  Is  clever  visualiza9on  the  key  to  hypothesis   forma9on,  even  if  “simple”  variables?   TED  talk  by  Deb  Roy  from  MIT  
  5. 5. 4.    When  do  you  give  up?  •  Reliance  on  visual  paGern  recogni9on  by   human  coder  may  not  reveal  relevant   (informa9ve)  features  (sound  spectrogram   cannot  be  “read”)  •  Failure  at  macro  level  prompts  search  for  info   at  micro  level  (fMRI  univariate  vs.  mul9variate   analysis):  need  to  “drill  down”  •  Failure  at  micro  level  may  indicate   indeterminacy  of  causal  hierarchy  (Fodor)  
  6. 6. 5.    Rules  of  sharing  •  When  does  “your”  data  become  accessible  by:   –  Your  collaborators   –  Friends  who  ask   –  Strangers   –  Anyone  •  Who  gets  credit?  •  How  should  junior  researchers  “share”?     Especially  with  senior  labs  that  have  $$$.  

×