Scribd
Upload
49 views7 pages

Practical 3 - Cropyield Forecasting - Exercise 5

This document provides instructions for using regression analysis in the Cropyield Statistical Tool (CST) to forecast crop yields. It explains how to select indicators and time periods, perform single indicator and best subset regression, and interpret the output. Exercises guide the user through practicing single indicator regression for maize in different regions and periods to identify the best predictors and their predictive power based on r-squared and RMSE values. Best subset regression is also demonstrated to potentially identify models using multiple indicators that further improve predictions.

Uploaded by

Dercio
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
49 views7 pages

Practical 3 - Cropyield Forecasting - Exercise 5

This document provides instructions for using regression analysis in the Cropyield Statistical Tool (CST) to forecast crop yields. It explains how to select indicators and time periods, perform single indicator and best subset regression, and interpret the output. Exercises guide the user through practicing single indicator regression for maize in different regions and periods to identify the best predictors and their predictive power based on r-squared and RMSE values. Best subset regression is also demonstrated to potentially identify models using multiple indicators that further improve predictions.

Uploaded by

Dercio
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 7

Cropyield Forecasting Pratical

Exercise 5: Regression
Analysis with CST

Por: Lucas Mauro Rogerio Chiau


Exercise 5 – Regression analysis with
CST

1.0

Objective........................................................................................................................2
2.0 Indicators tab...........................................................................................................2
3.0Options tabY.............................................................................................................4

1.0 Objective
The objective of this exercise is to use the regression analysis of the CST to
forecastyield.

2.0 Indicators tab


After selecting the “Regression analysis” tab, the first stage is to select indicators
from the “Available indicators” selection box.
Figure 1: The indicators for yield forecasting.

Exercise 1: Indicator Selection


Select region “Eastern Cape”, crop total maize and period J-III (third dekad of June).
Select the regression tab.
Q1: How many indicators are available in the “Available indicators” selection box?
Now select the period first dekad of February (F-I).

Q2: How many indicators are now available in the “Available indicators” selection box.
Which indicator is now available and why?

Now select the period: first dekad of April (A-I).


Q3: How many indicators are now available in the “Available indicators” selection box?
Which indicator is now available and why?

Q4: Some indicators are listed as having missing values. Particularly the satellite indicators
(SPOT-VGT) have many missing values. Can you explain this? Tip: have a look at the “Data
view” window in the trend analysis.
Add all indicators in the “available indicators selection box” into the box with “Free
indicators” using the >>button. When adding all indicators, the CgmsStatTool will
come up with the following window:

This is needed because all indicators and official yield need to be present in the
analysis. Press the “Synchronize” button to have all indicators synchronized over the
same years. To see what happened check the Time trend analysis page.

The CgmsStatTool can also compute the correlation matrix between the different
indicators. An indicator that is highly correlated (large absolute coefficient of
correlation) with another indicator has little additional value in the regression
analysis. A quick look on the correlation matrix will show you that all CGMS indicators
(01, 06, 08) are highly correlated. The sum of rainfall (05) and satellite derived
indicators (01, 02) are much less correlated.

3.0 Options tab

The options tab lists a number of options for regression models (see Figure 3). The
most fundamental choice to make here is to use either “Single free indicator” or “Best
subset selection”. In the first case only one indicator at a time is used to build the
regression model, in the second case the CgmsStatTool will test many combinations
of indicators to find the best fitting model.
Figure 3: Explanation of functionality in the regression options window
Exercise 2: Single Free indicator analysis
Settings in the time trend analysis window:
 Region: Limpopo
 Crop: Total maize
 period: June-I (first decade of June)
 Start/End year: 1987-2015
 Continue in the Regression Tab:
 Add all indicator to the Free Indicators

 Continue in the Options Tab:


 Select “Single Free indicator”
 Mark all indicators as having a positive sign
 Leave the rest default

Continue to the Output Tab


The table shows you the results of the regression models for the different
indicators.
Q1: How many regression models are listed and what is the Nonemodel?
Q2: What is the best model and what is the R values of this model?
2

Q3: A measure of the error on the crop yield prediction can be obtained from the
Root Mean Squared Error (RMSE) for prediction. What is the RMSE for the best
model and how does it compare to the Nonemodel?
What is the predicted yield?
Q4: Some of the cells with the t-values are marked either orange or red, what
does this mean?
Task2: Single Free indicator analysis
Repeat exercise 2, but set the period to April-III
Q1: What is the best model and what is the R values of this model?
2

Q2: What is the RMSE for the best model and how does it compare to the
Nonemodel.
Q3: What is the predicted crop yield (Prediction for target year)?
Q4: Is the predicted yield very different from the predicted yield in March-I

We will now repeat exercise 2 with the “Best subset selection”. Using the “Best
subset selection” we can try to build a better regression model and obtain a better
forecast with smaller error bounds.

Exercise 3: Best subset selection


Repeat exercise 2, (set the period to April-III). Continue in the Options Tab:
 Select “Best subset selection”
 Mark all indicators as having a positive sign
 Leave the remaining settings default

Continue to the Output Tab


Q1: What is difference in the ordering of models and why?
Q2: What is the best model with a single indicator? How does this compare to
the result you found with Question 1 of Exercise 2?
Q3: Is there a better model with two indicators, with three, with four?

TASK: Repeat exercise 3 above but now using yellow maize and FEB-II

You might also like