Вы находитесь на странице: 1из 27

Control of Experimental Error

Bulls eye represents the true value


of the parameter you wish to estimate

Accuracy = without bias


average is on the bulls-eye
achieved through randomization
Precision = repeatability
measurements are close together
achieved through replication

Both accuracy and precision are needed!


Randomization
To eliminate bias
To ensure independence among observations
Required for valid significance tests and interval estimates

Low High

Old New Old New Old New Old New

In each pair of plots, although replicated, the new variety is


consistently assigned to the plot with the higher fertility level.
Replication
The repetition of a treatment in an experiment

A B D A

C D B C
B A D C
Replication
Each treatment is applied independently to two or more
experimental units
Variation among plots treated alike can be measured
Increases precision - as n increases, error decreases

Standard error Sample variance


of a mean Number of replications

Broadens the base for making inferences


Smaller differences can be detected
Effect of number of replicates

Effect of replication on variance


8.0
7.5
Variance of the mean

7.0
6.5
6.0
5.5
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0

0 5 10 15 20 25 30 35 40 45 50

number of replicates
What determines the number of replications?

Pattern and magnitude of variability in the soils


Number of treatments
Size of the difference to be detected
Required significance level
Amount of resources that can be devoted to the
experiment
Limitations in cost, labor, time, and so on
Strategies to Control Experimental Error
Select appropriate experimental units
Increase the size of the experiment to gain more
degrees of freedom
more replicates or more treatments
caution error variance will increase as more heterogeneous
material is used - may be self-defeating
Select appropriate treatments
factorial combinations result in hidden replications and therefore
will increase n
Blocking
Refine the experimental technique
Measure a concomitant variable
covariance analysis can sometimes reduce error variance
The Field Plot
The experimental unit: the vehicle for evaluating
the response of the material to the treatment
Shapes
Rectangular is most common - run the long dimension parallel to
any gradient
Fan-shaped may be useful when studying densities
Shape may be determined by the machinery or irrigation
Plot Shape and Orientation
Long narrow plots are preferred
usually more economical for field operations
all plots are exposed to the same conditions
If there is a gradient - the longest plot dimension should
be in the direction of the greatest variability



Border Effects
Plants along the edges of plots often perform differently
than those in the center of the plot
Border rows on the edge of a field or end of a plot have an
advantage less competition for resources
Plants on the perimeter of the plot can be influenced by
plant height or competition from adjacent plots
Machinery can drag the effects of one treatment into the
next plot
Fertilizer or irrigation can move from one plot to the next
Impact of border effect is greater with very small plots
Effects of competition
In general, experimental materials should be evaluated
under conditions that represent the target production
environment


Minimizing Border Effects
Leave alleys between plots to minimize drag
Remove plot edges and measure yield only on
center portion
Plant border plots surrounding the experiment
Experimental Design
An Experimental Design is a plan for the assignment
of the treatments to the plots in the experiment
Designs differ primarily in the way the plots are
grouped before the treatments are applied
How much restriction is imposed on the random
assignment of treatments to the plots

A B D A A B D C

C D B C C D B A

B A D C B A D C
Why do I need a design?
To provide an estimate of experimental error
To increase precision (blocking)
To provide information needed to perform tests
of significance and construct interval estimates
To facilitate the application of treatments -
particularly cultural operations
Factors to be Considered
Physical and topographic features
Soil variability
Number and nature of treatments
Experimental material (crop, animal, pathogen, etc.)
Duration of the experiment
Machinery to be used
Size of the difference to be detected
Significance level to be used
Experimental resources
Cost (money, time, personnel)
Cardinal Rule:

Choose the simplest


experimental design
that will give the
required precision
within the limits of the
available resources
Completely Randomized Design (CRD)

Simplest and least restrictive


Every plot is equally likely to be assigned to any
treatment

A B D A

C D B C

B A D C
Advantages of a CRD
Flexibility
Any number of treatments and any number of
replications
Dont have to have the same number of replications
per treatment (but more efficient if you do)
Simple statistical analysis
Even if you have unequal replication
Missing plots do not complicate the analysis
Maximum error degrees of freedom
Disadvantage of CRD
Low precision if the plots are not uniform

A B D A

C D B C

B A D C
Uses for the CRD
If the experimental site is relatively uniform
If a large fraction of the plots may not respond or
may be lost
If the number of plots is limited
Design Construction
No restriction on the assignment of treatments to the
plots
Each treatment is equally likely to be assigned to any
plot
Should use some sort of mechanical procedure to
prevent personal bias
Assignment of random numbers may be by:
lot (draw a number )
computer assignment
using a random number table
Random Assignment by Lot
We have an experiment to test three varieties:
the top line from Oregon, Washington, and Idaho
to find which grows best in our area ----- t=3, r=4

A1 15
12
6
2 3 4
A A
5 6 7 8
A
9 10 11 12
Random Assignment by Computer (Excel)

In Excel, type 1 in cell A1, 2 in


A2. Block cells A1 and A2. Use
the fill handle to drag down
through A12 - or through the
number of total plots in your
experiment.
In cell B1, type = RAND(); copy
cell B1 and paste to cells B2
through B12 - or Bn.
Block cells B1 - B12 or Bn, Copy;
From Edit menu choose Paste
special and select values
(otherwise the values of the
random numbers will continue to
change)
Random numbers in Excel (contd.)
Sort columns A and B
(A1..B12) by column
B
Assign the first
treatment to the first r
(4) cells in column C,
the second treatment
to the second r (4)
cells, etc.
Re-sort columns A B
C by A if desired.
(A1..C12)
Rounding and Reporting Numbers
To reduce measurement error:
Standardize the way that you collect data and try to be as
consistent as possible
Actual measurements are better than subjective readings
Minimize the necessity to recopy original data
Avoid rekeying data for electronic data processing
Most software has ways of importing data files so that you dont
have to manually enter the data again

When collecting data - examine out-of-line figures


immediately and recheck
Significant Digits
Round means to the decimal place corresponding to
1/10th of the standard error (ASA recommendation)
Take measurements to the same, or greater level of
precision
Maintain precision in calculations

If the standard error of a mean is 6.96 grams, then


6.96/10 = 0.696 round means to the nearest 1/10th gram
for example, 74.263 74.3
But if the standard error of a mean is 25.6 grams, then
25.6/10 = 2.56 round means to the closest gram
for example, 74.263 74
Rounding in ANOVA
In doing an ANOVA, it is best to carry the full number of
figures obtained from the uncorrected sum of squares

If, for example, the original data contain one


decimal, the sum of squares will contain two
places
2.2 * 2.2 = 4.84

Do not round closer than this until reporting final results