REGRESSION
ANKITHA CS
❑ SPECIAL NAMES OF THE PARAMETERS ‘𝒂’ AND ‘𝒃’ AND
EXPRESSION FOR THEIR VARIANCES
▪ Regression equation is 𝒀 = 𝒂 + 𝒃𝑿
▪ The parameter ‘𝒂’ is called the 𝒀 intercept .It is the value 𝑌 assumes
when 𝑋 = 0
▪ The parameter ‘𝒃’ is called the regression coefficient and gives the
slope of the regression line,
▪ It shows how steep the line is
▪ The regression coefficient indicates the rate of change in the
dependent variable per unit change in the independent variable.
The variance of the estimates 𝑏 and 𝑎 are respectively given by
2 1 𝑆𝑦 2
𝑆𝑏 = − 𝑏2
𝑛−2 𝑆𝑥
2 2 𝑛−1
𝑆𝑎 = 𝑆𝑏 × 𝑆𝑥 2 + 𝑥ҧ 2
𝑛
Where 𝑆𝑦 and 𝑆𝑥 are standard deviations of 𝑦 and 𝑥 respectively
2 𝑛 σ 𝑦2− σ 𝑦 2 2 𝑛 σ 𝑥2− σ 𝑥 2
𝑆𝑦 = and 𝑆𝑥 =
𝑛(𝑛−1) 𝑛(𝑛−1)
❑ VARIANCE ABOUT THE REGRESSION LINE
(DEVIATION FROM REGRESSION)
▪ The assumption behind the standard linear regression is that each 𝑌𝑖
is normally distributed with mean value 𝑎 + 𝑏𝑋𝑖 and with constant
variance 𝜎 2 which is not dependent on the value of 𝑋𝑖 .
▪ The formula for estimate of this variance is given by
1
𝑆2 = σ 𝑌𝑖 − 𝑎 − 𝑏𝑋𝑖 2
𝑛−2
▪ The convenient formula to work out this variance is given by
𝟐 (𝒏−𝟏) 𝟐 𝟐 𝟐
𝑺 = 𝑺𝒚 − 𝒃 𝑺𝒙
(𝒏−𝟐)
❑ ASSUMPTIONS OF SIMPLE LINEAR REGRESSION ANALYSIS
▪ Relationship between 𝑋 and 𝑌 is linear
▪ Independent variables measured without error (fixed 𝑋)
▪ Errors are independent
▪ Errors are normally distributed
▪ Errors have constant variance 𝜎 2
➢ APPLICATIONS OF LINEAR REGRESSION ANALYSIS
1) Conversion between different length measurements i.e., from say
total length to standard length or to fork length and so on.
2) Calculation of fish lengths from scale measurements
3) Fitting weight-length relationship
4) Estimating parameters of the von Bertalanffy growth curve
5) Estimation of heritability of different traits based on the regression
of parents on offspring's
PROBLEMS
EXAMPLE 1:-
The data on fish yield tested under 5 stocking densities are given below
Sr.NO 1 2 3 4 5
Fingerlings(‘000/ha) 2 3 4 5 6
Fish yield(t/ha) 2.5 3.6 4.4 5.0 5.4
Compute:
i. The regression equation of fish yield on stocking density.
ii. Estimate fish yield for stocking density of 4,200/ha
iii. Mean square due to deviations from regression
Solution:-
Fingerlings Fish yield
Sr.NO (‘000/ha) (t/ha) 𝐗𝐘 𝐗𝟐 𝐘𝟐
𝐗 𝐘
1 2 2.5 5.0 4 6.25
2 3 3.6 10.8 9 12.96
3 4 4.4 17.6 16 19.36
4 5 5.0 25.0 25 25.10
5 6 5.4 32.4 36 29.16
Total 20 20.9 90.8 90 92.83
i) 𝑏 = 0.72
𝑎 = 1.3
and regression equation is 𝑌 = 1.3 + 0.72𝑋
ii) To obtain the estimate of fish yield when the stocking density is 4,200
Put 𝑋 = 4.2 in the above equation
we get
𝑌 = 1.3 + 0.72𝑋
= 1.3 + (0.72 × 4.2)
= 4.32 𝑡𝑜𝑛𝑠
iii) Mean square due to deviations from regression is given by
(𝒏−𝟏)
𝟐
𝑺 = 𝑺𝒚 𝟐 − 𝒃𝟐 𝑺𝒙 𝟐
(𝒏−𝟐)
2 𝑛 σ 𝑦2− σ 𝑦 2 2 𝑛 σ 𝑥2− σ 𝑥 2
𝑆𝑦 = 𝑆𝑥 =
𝑛(𝑛−1) 𝑛(𝑛−1)
5×92.83 −(20.9)2 5×90 −(20)2
= =
5×4 5×4
= 1.367 = 2.5
(𝒏−𝟏)
𝟐
𝑺 = 𝑺𝒚 𝟐 − 𝒃𝟐 𝑺𝒙 𝟐
(𝒏−𝟐)
2 4
∴𝑆 = 1.367 − 0.72 2 (2.5)
3
= 1.333 × 0.071
= 0.0947
➢ IMPORTANT USES OF STANDARD ERROR
Standard error plays an important role in statistical theory . Following are its
important uses:
▪ To measure the precision of a statistics . Higher the standard error lower is the
precision of a statistic
▪ To fix confidence limits for population parameters
▪ To determine the size of the sample required to achieve the desired precision
▪ To compute test statistic in tests of significance
➢ DIFFERENCE BETWEEN CORRELATION AND REGRESSION
CORRELATION REGRESSION
Analysis of relationship between two or Measure of average relationship
more variable between two or more variable
Range of correlation coefficient −1 𝑡𝑜 1 Range of regression coefficient is
− ∞ 𝑡𝑜 ∞
There is no difference in dependent and Both variables are different
independent variables
Correlation coefficient is not affected by Regression coefficient is affected by
change in origin and scale change in origin and scale
If value of one variable is known, then If the value of one variable is known,
the value of other variable cannot then the value of other variable can be
estimated estimated
