Mas Maiz: Chart historical corn yields with proc transreg

[#1]

Fans of Reggaeton may know the 2006 N.O.R.E hit ‘Mas Maiz,’ which translates literally from Spanish as ‘More Corn’. If you wanted to understand just HOW MUCH more corn, you can use proc transreg on data published from USDA National Agricultural Statistics Service.

Optimists would predict that corn yields tend to increase over time. While that’s true, there were also epochs of exponential yield increase. Other pundits have grouped these epochs, and I’ll call them:

1866-1936 “The hardscrabble age”

1937-1955 “The hybrid age”

1956-1995 “The age of genetics and pesticides”

1996-2019 “The biotech age”

We not only want to see if there is a relationship between year and yield, but also to break up the years into logical clusters, and perform regression analyses on each of them.  So, we will use proc tranreg as follows:

ods graphics on;
proc transreg data=import1 ss2 plots=fit(nocli noclm);
   ods output coef=coef;
   model identity('US Corn Yield (Bu/A)'n) = class(Group / zero=none) | identity(year_);
run;

To perform a regression and ANOVA, we used the ss2 option in the proc line.

To break up the years into groups of interest, we included a column named ‘Group’ in the dataset to classify the groups, and then specify to model the response variable by ‘class(group)’.

After running the model, we see the overall regression is significant, with an adjusted R2 of 0.9792 (not shown). One of the largest deviants was year 2012, which was a major drought year throughout the Midwest US. And the regression coefficients bear out that corn yields were unpredictable and non-increasing, until the hybrid age dawned around 1937.

The regression table is:

Variable	DF	Coefficient	Type II Sum of Squares	Mean Square	F Value	Pr > F	Label
Class.Group1866-1936	1	24.7429090	4679.7	4679.7	97.22	<.0001	Group 1866- 1936
Class.Group1937-1955	1	45.7953938	5290.0	5290.0	109.90	<.0001	Group 1937- 1955
Class.Group1956-1995	1	60.4673001	52172.5	52172.5	1083.86	<.0001	Group 1956- 1995
Class.Group1996-2019	1	57.7120584	1662.3	1662.3	34.53	<.0001	Group 1996- 2019
Identity(Group1866- 1936Year_)	1	-0.0000501	10.0	10.0	0.21	0.6495	Group 1866- 1936 * Year_
Identity(Group1937- 1955Year_)	1	0.0020928	333.0	333.0	6.92	0.0094	Group 1937- 1955 * Year_
Identity(Group1956- 1995Year_)	1	0.0051000	18495.1	18495.1	384.23	<.0001	Group 1956- 1995 * Year_
Identity(Group1996- 2019Year_)	1	0.0053416	4377.5	4377.5	90.94	<.0001	Group 1996- 2019 * Year_

The real question is: are yields going to continue to increase? Or get more predictable? As precision agriculture and biologicals are gaining a foothold, are we entering a new age? Perhaps that is a forecasting question for another day!