Title: Plant Science 547
1Plant Science 547 Biometrics for Plant Scientists
Association Between Characters
2Effect of One Treatment on Another
- Test hypothetical models for biological systems.
To explain relationships (i.e. linear, quadratic,
etc., orthogonal contrasts). - To predict the values of one variable according
to set values of another.
3Possible Relationships of Interest
- Predict optimal nitrogen application to maximize
seed yield. - Determine deficiencies in national supply of
specific agricultural products by relating yield
to weather related characters (rainfall,
sunshine, etc). - Explain relationship between plant biomass and
time after seeding to select for more insect
tolerant cultivars.
4- Charles Darwin
- Born in England, educated in Scotland.
- The father of evolution
- Most famous for his travels on the Beagle to the
Galapagos Islands. - Survival of the fittest.
- Wrote The Origin of Species.
5History
- 19th Century - Charles Darwin.
- Francis Galton In the law of universal
regression each peculiarity in a man is
shared by his kinsman, but on average to a lesser
degree. - Karl Peterson Andrew Lee (statisticians) survey
1000 fathers and sons height. - Using this data set Galton, Peterson and Lee
formulated regression analyses.
6Regression Models
- Dependant variable (of interest). Usually donated
by y - One, or more, independent variable on which the
dependant variable is related in a specific
manner. Usually donated by x, x1, x2, etc.
7Common Types of Regression
- Simple linear regression
yb0b1x - Non-linear regression yb0b1xb2x2 yex
yln(x) - Multiple regression yb0b1x1b2x2
8Nitrogen application v Seed yield
9Nitrogen application v Seed yield
b1
bo
Y bo b1x
10Simple Linear Regression
Y bo b1x
b1 SP(x,y)/SS(x)
SP(x,y) ?(xi-x)(yi-y)
SP(x,y) ?(xy) - ?(x) ?(y)/n
SS(x) ?(xi-x)2 SS(x) ?(x2) - ?(x)2/n
11Simple Linear Regression
Y bo b1x
bo mean(y) b1 x mean(x)
12Linear regression example
- Sex-linked mutations in Drosophila.
- x-variable is dosage of radiation (1000s rads).
- y-variable is the percentage of mutation observed
in Drosphila populations.
13Mutation Frequency in Drosophila
14Linear Regression Example
15Linear Regression Example
SS(x) ?(xi2)-?(xi)2/n 39.25 - (11.5)2/5 12.800
Mean (y) ?(yi)/n 11.5/5 2.30
16Linear Regression Example
SS(x,y) ?(xiyi)-?(xi) ?(yi)/n 89.35 - (11.5 x
26.3)/5 28.860
b1 SP(x,y)/SS(x) 28.860/12.800 2.255 b0 y
- b1x 5.26 - 2.255 x 2.30 0.735
17Mutation Frequency in Drosophila
Y 0.0735 2.255 x
18Analysis of Variance Regression
- Total variation of the dependant variable (the
one of interest). - Partition into variation accountable by the
regression model (linear or other) Sum of
squares for regression. - Other, non-explaiable variation
Residual sum of squares.
19Linear Regression Example
Total SS SS(y) ?(yi2)-?(yi)2/n 204.11 -
(26.3)2/5 65.772
Regression SS SP(x,y)2/SS(x) 28.8602/12.800
65.070
Residual SS ?2Res Total SS - Regression
SS 65.772 - 65.070 0.702
20Analysis of Variance Regression
Residual can be tested if observations are
replicated
21t-test and regression
22t-tests and Regression
Is the regression slope significantly greater
than zero?
t b-0/se(b) b/se(b)
se(b) ?SS(y) - b1 SP(x,y)/(n-2)SS(x)
?65.772 - 2.255 x 28.860/(3 x 12.800
0.134
23t-tests and Regression
Things to Note
se(b) ?SS(y) - b1 SP(x,y)/(n-2)SS(x)
SS(y) - b1 SP(x,y) Residual SS
SS(y) - b1 SP(x,y)/(n-2) Residual MSq
se(b) ?Residual MSq/SS(x) ??2Res/SS(x)
24t-tests and Regression
Is the intercept significantly different from a?
t b0-a/se(b0)
se(b0) ??2Res . 1/n mean(x)2/SS(x)
?0.234 x 1/5(2.32/12.800 0.378
25Predicting the dependant variable!
26Predicting the Dependant Variable
Y 0.0735 2.255 x
At x 2.5 - 1000 Rads y 0.073 2.255 x 2.5
5.7105
How accurate is this estimation?
-
se(yp) ??2 11/n(xp-x)2/SS(x)
27Linear Regression Example
Predicting at x 2.5 se(yp) ?0.234
11/5(2.5-2.3)2/12.800 se(yp) 0.531 yp
5.7105 0.531
28Linear Regression Example
Predicting at x 4.5 se(yp) ?0.234
11/5(4.5-2.3)2/12.800 se(yp) 0.663 yp
10.2205 0.663
29Mutation Frequency in Drosophila
30(No Transcript)