The Use of Statistics in Crop Management

1
The Use of Statistics in Crop Management
2
Management

• The act, manner, or practice of handling, or
  controlling something
• To direct, supervise, or carry on business
  affairs
• Making and implementing a decision

3
Critical Thinking

• Characterized by careful and exact evaluation and
  judgement (American Heritage Dictionary)

4
Why, When, and How?(should we do something)
5
Why ?

• Why should I plant this crop?
• Why should I plant this variety?
• Why should I use this product?
• Why is the tillage system important?

6
When?

• When should I plant?
• When should I harvest?
• When should I spray?
• When should I apply nutrients?

7
How?

• How do I maximize profit?
• How do I minimize the environmental impact?
• How do I manage risk?
• How do I know when to do what in crop management?
• How does this product perform?
• How does its performance compare to other
  products?
• How much does it cost vs. the expected return?

8
Rules 1) Draw 4 straight lines that intersect
all of the dots without lifting your pencil. 2)
Two minutes.
9
Rules 1) Draw 4 straight lines that intersect
all of the dots without lifting your pencil. 2)
Two minutes.
10
Biased Information

• Self-imposed limits on thinking
• Improper techniques involving information
  collection
• Improper interpretation of results
• Biased source/personal agenda or filter
• Everyone has an opinion and everyone has an
  agenda how do we sort through opinions and
  agendas?

11
Roles in Agriculture

• Regulator
• NRCS
• DENR
• EPA

• Adviser
• Extension Agent
• NCDA Agronomist
• NRCS
• Consultant
• Agribusiness Support and Sales
• Farm Manager/Owner

12
Importance of Objectivity

• The best decisions are based on the best possible
  information
• Objectivity is critical in effective advising,
  management, and regulation

13
Decision Making Based on Data Analysis

• Questions to ask
• Where was the location?
• How many locations?
• How many years of data?
• Who conducted the test?
• Was it statistically analyzed?
• Are all of the data shown?
• Is it a test or a testimonial?

14
An Experiment is

• A planned inquiry to obtain new facts or to
  confirm or deny results of previous experiments
  (Steele and Torrie)
• Results will be used in making a decision

15
Experiments vs. Observational Studies

• Controlled Experiment Experimental Units
  (treatments) are assigned randomly under
  controlled conditions in a manner to define cause
  and effect relationships in order to keep factors
  other than treatment constant
• Observational Study Observe a selected
  population and record what you see (provides a
  report of observations)

16
Agricultural Applications of Statistical Analysis

• The basic purpose of statistical analysis is to
  measure variability in observations across an
  experiment and to assign that variability to
  known effects (treatment and replication) and
  unknown effects (error).
• A high ratio of variability from known sources to
  unknown sources is required to conclude that
  observed differences are due to treatments and
  not some other uncontrolled or unknown effects.
• This process allows the researcher to have
  confidence that the differences observed are due
  to treatment and not due to environment or other
  unknown causes.

17
Experimental Design

• Randomization All plots have an equal chance of
  being assigned a given treatment and are assured
  unbiased estimates of treatment means and
  experimental error
• Replication Improves precision of treatment
  means and is a measure of consistency of response
  (repeatability)
• -More replication greater precision

18
Experimental Design

• Local Control (Blocking) Plots are grouped into
  blocks with similar features (soil type, texture,
  OM, slope), but features between blocks are
  often different thereby improving precision by
  accounting for a portion of the variation

19
Weed control
0
75
50
90
0
85
100
90
0
100
100
100
0
85
100
100
20
Weed control
heavy
Rate 2
Check
Rate 1
Rate 3
4
0
75
50
90
Check
Rate 1
Rate 2
Rate 3
3
0
85
100
90
Moderate
Check
Rate 1
Rate 2
Rate 3
2
0
100
100
100
Check
Rate 1
Rate 3
Rate 2
1
0
85
100
100
light
21
Separating (Partitioning) Variability into Known
and Unknown Sources

• A common procedure used to determine the causes
  of observed variability is called the Analysis of
  Variance (ANOVA).
• The ANOVA determines if a significant portion of
  the observed variation is due to treatment. But,
  the general ANOVA does not determine differences
  among treatments.
• Multiple comparison procedures, contrasts, and
  regression are used to separate differences among
  treatments.
• Often times more can be concluded from the ANOVA
  table than from a table of means or a graph
  (relationships are important)

22
Hypothesis Testing Statistician Terms-Null
hypothesis no difference in populations-If
reject null hypothesis, then a difference exists
among at least two of the populations being
compared
23
Hypothesis Testing Peanut Planting Date
Experiment-Null hypothesis no difference in
yield regardless of planting date-If reject null
hypothesis, then a difference in yield can be
attributed to planting date
24
Hypothesis Testing Peanut Planting Date
Experiment-Null hypothesis no difference in
yield regardless of planting date-If reject null
hypothesis, then a difference in yield can be
attributed to planting date
25
ANOVA
26
Assumptions

• Normal distributions (bell-shaped curve)
• Appropriate controls in place

27
Statistical Tools

• Mean separations
• Regressions
• Correlations
• Factorial treatment arrangements
• Split plot designs vs. randomized complete block
  designs

28
Examples

• Mean separation
• Correlations
• Regression

29
Step 1 Digging and inverting Step 2 Harvest
(combining)
30
Determining Pod Maturity of Peanut
The indeterminate nature of peanut contributes to
the challenge of deciding when to dig and invert
vines in order to optimize pod yield and market
grade characteristics
Progression of maturity of peanut pods using pod
mesocarp color
31
We tell the farmer when we think the peanuts will
be at optimum maturity. The farmer decides when
the peanuts are ready to dig.
32
Example 1Objectives

• Is there a relationship between canopy
  reflectance and pod maturity?
• Null hypothesis there is no relationship
  between canopy reflectance and pod maturity
• Does planting date affect pod yield?
• Null hypothesis peanut yield is the same
  regardless of planting date

33
Example 1Materials and Methods

• Planted NC-V 11 in one trial and VA 98R in a
  different trial
• Randomized complete block with 4 replications,
  repeated over time
• Planted VA 98R on four different dates and NC-V
  11 on two different dates
• Determined the percentage of mature pods (MP)
  and canopy reflectance (multi-spectral imaging)
  on one date in mid September
• Using ANOVA, mean separation, and correlations to
  approach answers to the objectives

34
Example 1Objective

• Does planting date affect pod yield?
• Null hypothesis peanut yield is the same
  regardless of planting date
• Using ANOVA and mean separations to determine if
  yield differed as a result of planting date

35
(No Transcript)
36
(No Transcript)
37
Summary VA 98R

• Variation in yield and the percentage of mature
  pods (MP) was noted when comparing planting
  dates and years
• Yield at the May 15-18 planting date was as high
  or higher than yield at the other planting dates
  regardless of year when peanut was dug at optimum
  maturity
• When the MP was determined on a single date in
  September, variation in MP was noted during all
  years.
• Generally, delaying planting resulted in a higher
  MP when determined on a single date in September

38
(No Transcript)
39
Summary Cultivar NC-V 11

• No variation in yield was noted when comparing
  planting dates when peanut was dug at optimum
  maturity
• A higher percentage of mature pods was noted on a
  single date in September when comparing planting
  dates

40
Example 1Objective

• Is there a relationship between canopy
  reflectance and pod maturity?
• Null hypothesis there is no relationship
  between canopy reflectance and pod maturity
• Using correlations to define relationships
  between canopy reflectance at various bandwidths
  with pod maturity

41
(No Transcript)
42
Abbreviated Summary Cultivars NC-V 11 and VA 98R

• Canopy reflectance at various band widths was
  negatively correlated with the MP for the
  cultivar VA 98R
• Canopy reflectance was not correlated with the
  MP for the cultivar NC-V 11
• More research s needed to determine potential of
  canopy reflectance as a consistent indicator of
  pod maturation

43
(No Transcript)
44
Example 2Objectives

• Do relationships exist among pod yield,
  percentage of extra large kernels (ELK), and
  percentage of total sound mature kernels (TSMK)?
• Null hypothesis there is no relationship
  among yield, ELK, or TSMK
• What affect does digging date have on pod yield,
  ELK, and TSMK?
• Null hypothesis there is no difference in
  yield, ELK, or TSMK regardless of digging date

45
Example 2Materials and Methods

• Dug the cultivar Gregory beginning in mid
  September through mid October on approximately
  weekly intervals
• Determined pod yield, ELK and TSMK for each
  digging date.
• Randomized complete block with 4 replications
• Using regression and correlations to approach
  answers to the objectives

46
Example 2Objectives

• What affect does digging date have on pod yield,
  ELK, and TSMK?
• Null hypothesis there is no difference in
  yield, ELK, or TSMK regardless of digging date
• Regressions testing linear, quadratic, and cubic
  functions tested yield, ELK, and TSMK versus
  days after peanut emergence

47
Percentage of error explained by DAP
Y -12521x 89x2 0.2x3 583517
48
(No Transcript)
49
(No Transcript)
50
Influence of days after peanut emergence on pod
yield, percent total sound mature kernels
(TSMK), and extra large kernels (ELK)Data are
presented as percent of maximum for each parameter
Percent of maximum yield
Days after peanut emergence
51
Example 2Objectives

• Do relationships exist among pod yield,
  percentage of extra large kernels (ELK), and
  percentage of total sound mature kernels (TSMK)?
• Null hypothesis there is no relationship
  among yield, ELK, or TSMK
• Used correlations to define significance of
  relationships among yield, ELK, and TSMK

52
(No Transcript)
53
Summary Digging Date and Gregory

• A significant polynomial relationship was
  observed for pod yield versus days after peanut
  emergence. Pod yield increased as digging was
  delayed up to approximately 154 days after
  emergence, and then yield decline with later
  digging dates.
• A significant linear relationship was noted for
  ELK and TSMK versus days after peanut
  emergence. The ELK and TSMK increased as
  digging was delayed.
• ELK and TSMK were positively correlated over
  the range of digging dates evaluated in this
  experiment.
• Yield and ELK or TSMK were not significantly
  correlated over the entire range of digging dates
  in this experiment.

54
Influence of days after peanut emergence on pod
yield, percent total sound mature kernels
(TSMK), and extra large kernels (ELK)Data are
presented as percent of maximum for each parameter
Percent of maximum yield
A significant correlation may exist between 134
and 154 DAE
Days after peanut emergence
55
Influence of days after peanut emergence on pod
yield, percent total sound mature kernels
(TSMK), and extra large kernels (ELK)Data are
presented as percent of maximum for each parameter
Percent of maximum yield
What are the dangers of extrapolating beyond the
data or excluding certain portions of the data?
Days after peanut emergence
56
Precision of Comparisons Versus Logistical
Constraints

• Randomized Complete Block Designs
• Split Plot Designs
• Split Block Designs
• Splitting Fields in Half (Strips)
• Comparing Different Fields

Partitioning experimental error and treatment
effects how can this be achieved given
logistical constraints?
57
Randomized Compete Block Design
58
Split Plot Design
59
Split Block Design
Rep 2
Rep 1
60
Breaking the Field into Strips
No replication and no estimate of variance
61
Comparing Fields
Ditch
Old grass waterway
Comparison of fields or sections of fields and
not really tillage and/or varieties
62
Precision of Comparisons Versus Logistical
Constraints

• Randomized Complete Block Designs
• Split Plot Designs
• Split Block Designs (variation?)
• Splitting Fields in Half (Strips) (variation?)
• Comparing Different Fields (variation?)

Partitioning experimental error and treatment
effects how can this be achieved given
logistical constraints?
63
(No Transcript)
64
The Use of Statistics in Crop Management

• Using statistics to make valid comparisons that
  can be extrapolated to other circumstances
• The most predictable and dependable
  recommendations include conclusions drawn from
  appropriately designed and analyzed experiments
  (regardless of the preconceived or expected
  outcome)

65
(No Transcript)
66
(No Transcript)
67
(No Transcript)
68
(No Transcript)