MPO 581 Class 3/27   Jan 26, 2011

Topic: Fields vs. lists; Teams, projects and roles

Loose ends from last class:

Today's material:Whiteboard -- about data sets

• We often have fields: 

• Multidimensional: var [lat, lon, z, t]
• Scale dimensions too: example var[lat, lon, z, month_of_year, year]
  
• Multivariate sets of the above: {var1, var2, var3}

• We sometimes have lists of events
  
• Example: hurricane database
• variables may be lat(events), lon(events), strength(events), size(events) 
• can build frequency (probability) fields and then treat with field methods
• can do some things directly: distributions, basic stats
  

• Structural vs. statistical dimensions
• Want results as a function of structural dimensions
• Want to summarize or transform statistical dimensions

• Kinds of statistical operations
• can do on a heap or pool of data:
  
• average, histogram, variance
• correlations and correlation matrix stuff
  
• lag correlation
  
• requires a full continuous sequence or series
• scale decomposition (Fourier/wavelet etc)

• Frequently encountered challenges with real data sets (yours??)
  
• too little, too much
• spotty/ missing
• randomly missing? or biased sampling?
  
• bounded domains (ocean basins, etc)
  
• strange distribution (like precip: a lot of zeros)
• not averageable (categorical, or uncalibrated,)
• have time want space (or vice versa)
• nonstationary errors
• inhomogeneous dimension (latitude, pressure, etc.) 
• scaling up from point to 'grid cell' / finite area

Team ideas (whiteboard; moving toward definiteness but ???) 

Open questions, assignments, and loose ends for next class:

Testable questions about today's material: