Transforming a sample design for taking into account new statistical needs, new information or new technological instruments for data collection

1
Transforming a sample design for taking into
account new statistical needs, new information or
new technological instruments for data collection
Elisabetta Carfagna Professor, University of
Bologna, Italy elisabetta.carfagna_at_unibo.it
2
Background and justification

• Various agricultural statistical systems
• New statistical needs
• New low cost technological instruments for data
  collection
• GPS
• PDAs
• New information is available (e.g. a land cover
  map produced for controlling subsidies)

3
Alternative strategies

• First strategy the sample design adopted in the
  previous years is preserved
• Second strategy the sample design is conceived
  ex novo
• Third strategy small changes are applied to the
  sample design

4
Italy, Agrit projectfrom early 90s to 2000

• From early 90s to 2000
• Stratified random sampling of about 5,000
  segments with physical boundaries
• Stratification land cover map from
  photo-interpretation of Landsat TM images

5
POPOLUS

• Availability of POPOLUS
• grid of points 500 meters apart
• photo-interpreted according to land cover legend
• geometrically corrected photos, scale 110,000
• POPOLUS first phase sample (used for
  stratification)

6
Agrit project from 2002

• From 2002
• Two phase stratified random sampling of 84,000
  un-clustered points
• CVs less than 5 for main crops at national level

7
Agrit new technological instruments

• From 2009 Ultra Mobile Personal Computer (UMPC) -
  PDA with GPS and GPRS/UMTS and Wi-Fi connection
• Data collected immediately sent to the central
  acquisition unit
• Change of quality control

8
Refresh project for control of declaration

• Availability of new information
• Refresh project for control of declaration of
  farmers (2009-2010)
• Photo-interpretation for the whole Italian
  territory of
• geometrically corrected aerial photos
• very high resolution satellite images with
  spatial resolution 0,5 m
• POPOLUS update

9
New information needs

• Baseline indicators for agro-environmental policy
  and rural development policy (EU)
• Agro-environmental parameters
• Farmland of high naturalistic value
• Risk of erosion

10
Observation on small circle around the point

• Observations on small circle around the point
• Irrigation
• Grass or bare soil (for permanent crops)

11
Observations on square segments

• Observations on square segments
• Irrigation, presence and kind
• Erosion, presence and level
• Bushes
• Small wood lots
• Isolated trees etc.

12
Pilot project

• Pilot project for
• Detecting the problems of the ground survey
• Evaluating the variance of the variables
• Computing sample size
• Identifying the optimum segment size

13
First strategy-preserving the sample design

• Advantages of preserving the sample design
• Perfect comparability over time
• Change detection
• Experience of the project managers and of the
  enumerators
• The reliability of the data collection procedures
  has already been assessed

14
First strategy-preserving the sample design

• Disadvantages of preserving the sample design
• New information (e.g. land cover map) cannot be
  used at the design level
• Only some new parameters can be estimated
• Cannot take complete advantage from the
  characteristics of new technological instruments

15
Second strategy-sample design conceived ex novo

• Advantages of conceiving the sample design ex
  novo
• New statistical needs satisfied
• New available information used at the design
  level
• Maximum advantage from new technological
  instruments for data collection

16
Second strategy-sample design conceived ex novo

• Disadvantages of conceiving the sample design ex
  novo
• Not complete comparability
• Difficult change detection
• Project managers and the enumerators have to
  adapt to new procedures
• Reliability of the data collection procedures has
  to be assessed
• Cost

17
Third strategy-small changes

• Advantages of small changes
• New statistical needs partially taken into
  account
• Comparability over time
• Experience of the project managers and of the
  enumerators
• Reliability of the data collection procedures

18
Third strategy-small changes

• Disadvantages of small changes
• New information (e.g. land cover map) cannot be
  used at the design level
• Few new parameters can be estimated
• Partial advantage from new technological
  instruments for data collection

19
Thank you for your attention elisabetta.carfagna_at_
unibo.it