More About:
Prior to the widespread adoption of allin, allout (AIAO) production systems, variation in growth was largely a “hidden” cost. Pigs were sorted from pens when they reached market weight, and the fact that some took longer than others went largely unnoticed, or at least ignored.
Furthermore, in continuousflow systems, downtime due to variable growth rates affects pen usage, while in AIAO systems it affects room or barn usage. Consequently, the economic impact of variation is much greater in AIAO systems.
Variation in market weights increases sort losses, a cost that goes straight to the bottom line. From a labor and management perspective, there is the annoyance factor of dealing with tailend pigs in a room or barn, which is also greater than in pens.
The cost of variation has always been with us, but those costs are now much more obvious.
In the past, the industry's singular focus was on “average” growth rate. Today, while we continue to pay attention to growth rate averages, greater attention is being focused on the “range” in that growth ? the variability of growth.
Measuring Variation
Statistically, variation can be defined in a variety of ways. The most common terms used to express it are standard deviation (SD) and coefficient of variation (CV), although range may also be useful.
Most people are familiar with the “bell curve,” which shows the typical distribution of measurements within a specific group. Many measurements can be described using a bell curve (weight, height, etc.).
If the measurement of a group or population is made and, when plotted, displays a bell shape, it is called a “normal” distribution. If the plotted data do not follow the bell shape, the data are called “skewed.”
For example, when the body weights of 632 pigs, averaging 20 weeks of age, are charted (Figure 1), it shows a very typical distribution of body weights for pigs of that age. This distribution is almost “normal,” but is skewed slightly to the left, reflecting the exaggerated number of tailenders in the group.
The shape of the bell curve reveals a great deal about a population or group. For example, if the bell shape is narrow, the population is relatively uniform because most of the measurements are closer to the average. If the bell shape is wide, the population is less uniform because more measurements are found further from the average.
One useful measure of the width of the bell curve is called the standard deviation (SD). The wider the bell shape, the larger the standard deviation, and the greater the variability of the group of animals.
Understanding the Terminology
Following are definitions of important statistics used to describe variation in a group of pigs.

Mean: The mean is the average of all weights within a group of pigs. It provides no indication of the variability of weights within the group.

Median: The median is determined by aligning all pig weights in order of magnitude (i.e., from smallest to largest or vice versa), then selecting the middle observation. If the distribution is “normal,” the median and the mean (average) will be very similar, if not identical.

Minimum, maximum and range: The minimum and maximum are selfevident; they are the lightest and heaviest weights within the group. The difference between the minimum and maximum is called the range. The wider the shape of the bell curve, thus, the less uniform the group of pigs, the larger will be the range.

Standard deviation (SD): The standard deviation is a measure of dispersion. The greater the variation in weight of a group of pigs, the larger will be the standard deviation.
In a “normal” distribution, statisticians have determined that one (1) standard deviation about the mean will include 68% of the pigs in the total group. Using the data in Table 1 for the 19dayold pigs, and assuming the data is distributed normally, the standard deviation has been calculated to be 2.7 lb., with a mean of 11.9 lb. Thus, we can estimate that 68% of the pigs, or 863 pigs in this group, will be within one standard deviation of the mean and weigh between 9.2 lb. (11.9  2.7) and 14.6 lb. (11.9 + 2.7).
Two standard deviations take in 95% of the pigs. Therefore, 1,206 pigs weigh between 6.5 lb. and 17.3 lb.
Three standard deviations will include 99% of the pigs, so 1,261 pigs weigh between 3.8 lb. and 20.0 lb. Of course, these numbers are approximations because data might not be perfectly, normally distributed, as Figure 1 shows, or because insufficient numbers of animals were weighed to adequately estimate these parameters.

Coefficient of variation (CV): The coefficient of variation is calculated by dividing the standard deviation by the mean, then multiplying by 100.
Referring again to the example in Table 1, the standard deviation of 2.7 lb. can also be presented as a coefficient of variation of 22.7%. The standard deviation becomes larger as the pigs grow.
Therefore, in order to determine if relative variation is increasing or decreasing as pigs grow, the coefficient of variation is often used.
In Table 1, while the standard deviation increases as the pigs become heavier, the coefficient of variation decreases, indicating that relative variability among pigs under commercial conditions typically declines as they grow.
There is one danger in using CV to represent variation. Sometimes, within pigs of the same age, the CV changes not because the SD changes, but because the mean weight changes (i.e., weaning weight increases). One could misinterpret this as meaning variation is reduced, when in actuality the variation did not change at all.
Measuring Variation On the Farm
There is surprisingly little information on normal distributions of bodyweights on commercial farms. Because it requires the weighing of many animals in order to accurately estimate SD and CV, it is not commonly done on most farms.
Furthermore, most research data cannot be used as reference points because in most experiments, animals are preselected to obtain a uniform group to increase experimental precision.
Geneticists have this information because variation is essential in selection programs, but even that data is not widely available.
Still, it is important that we develop an understanding of “normal” or “typical” variation, because it helps us to develop strategies for dealing with problems on individual farms or systems.
Standards can be used to determine if a given farm is better advised to manage variation or to minimize it. The difference between these two choices represents a critical management decision. The limited amount of data on the subject suggests that much larger variation is seen in some circumstances. CVs of 20% to 35% in growfinish pigs among commercial farms have been reported.
The minimum number of pigs that must be weighed in order to estimate a mean, a standard deviation or a coefficient of variation is not constant. Rather, the number depends on the intrinsic variability within the population.
At weaning, because variability is so high, weighing even 100 pigs provides a poor estimate of the CV. However, when pigs are removed from the nursery or at first pull in a finishing facility, randomly weighing as few as 50 pigs throughout the barn will provide satisfactory information.
Random selection for weighing is very important because weighing too many or too few of the lighter and heavier pigs within the group will skew the results. If too many “outliers” are weighed, variation will be overestimated; if too few are weighed, variation will be underestimated.
Using the weights of pigs at marketing does not provide an accurate estimate of variability because it represents a group of animals preselected according to their market weights. For this reason, we prefer to use the weights of pigs at first pull because all animals can be included. However, this is not a particularly convenient number to generate on most farms, so alternatives must be developed.
Causes of Variation
Many factors can affect the degree of variability observed on a given farm. Without question, a certain amount of variability is “programmed” at the time a pig is born. Pigs that are smaller at birth are compromised physiologically and socially, and their expected performance will fall short when compared to their heavier contemporaries.
In addition to these innate contributors to variability, there are many external forces that come into play as well, such as health status, access to resources (feed, water) and poor ventilation.

Prenatal influences: Variation begins on the day pigs are born. Studies of neonatal and weaned pig management have reported that even within a litter, the CV for birth weight is between 22% and 26%. As litter size increases, the average birth weight declines by about 0.1 lb. for every additional piglet in the litter.
While differences in birth weight are obvious to any farrowing technician, what is less clear is the relationship between birth weight and physiological “competency” at birth.
For example, differences associated with low birth weight have been observed in reduced number and height of intestinal villi, lactase and lipase activity, reduced muscle respiratory enzyme activity, fewer muscle thyroid hormone receptors, lower IGF1 levels in the blood and fewer muscle fibers.

Postnatal influences: After birth, additional factors contribute to variability. For example, heavier birth weight piglets consume about 30% more milk than their lighter littermates. In addition, heavier birth weight piglets, or at least those that win the most fights early in life, tend to suckle the anterior teats on the sow, which are known to deliver higher milk volumes. In one study, pigs nursing the anterior teats were 3.3 lb. heavier at weaning than those adopting the posterior teats.
Lower milk intake is not only associated with slower growth, but also reduced wholebody protein synthesis, according to research conducted in the United Kingdom. That research also reported that tripling milk intake during the first week of life quadrupled the pigs' protein deposition rate.

Postweaning influences: One of the most predictable contributors to variability in the postweaning period is the variability in weaning weight. For example, the correlation between weaning weight and nursery exit weight was found to be 0.73. Numerous authors have related weaning weight to nursery exit weight by suggesting that for every 1 lb. increase in weaning weight, nursery exit weights will increase by a given amount.
At the Prairie Swine Center Elstow Research Farm, we have found that for every 1 lb. increase in weaning weight, there is a 1.9 lb. increase in nursery exit weight (56 days of age) and a 4.2 lb increase in market weight. This relationship varies widely among farms.

Herd health and pathogen exposure: While inherent factors such as birth weight, weaning weight and suckling habits affect variability in pigs, there are many other “external” forces at play as well.
Perhaps the most important external force on variability is the disease status of the herd. Because the extent of disease exposure differs among animals, and because the impact of that exposure on animal health and performance also differs among individuals, it is not surprising to observe that herd health status can have a major impact on variability.

Feed and water: Access to resources, such as feed and water, is also a potential contributor to variability.
If feed or water access is limiting, dominant pigs in a pen will have an advantage over subordinate pigs, which results in greater growth disparity.
Social Behavior Model
There is a very interesting social behavior model for describing variability where one considers the average performance of a group of pigs along with the magnitude of the variation about this mean.
If average performance is very good and the degree of variation is low, one can assume that the conditions in which the pig are being reared are good and there is no depression in growth due to aggressive behaviors. Conversely, where performance is suboptimal and variation is high, one can assume there is competition for resources, such as food and water.
Overcrowding is not a likely issue in this scenario, because space is not a resource that pigs can hoard away from other pigs in the pen. However, pigs can prevent others from eating or drinking. The resultant uneven distribution of feed and water does lead to increased variability.
In the third scenario, performance is poor but variability is low, indicating an overall impairment due to a poor environment. Crowding could be the problem in this scenario, because crowding, unless severe, will uniformly depress growth, but not increase variability.
Addressing Variability
The challenge of variability can be addressed in two ways ? reducing variability and managing variability.
If variability within a barn is already quite low, producers are more likely to achieve success by seeking ways to manage variability. But if variability is high, then clearly there are deficiencies within the barn that are impairing performance and should be addressed. In this instance, it should be possible to reduce variability.
Reasonable targets for variability in a barn can be based on CV as described above. Although information on “normal” variability is admittedly limited, I would offer the following thresholds for CV:

20% of weaning weights;

12 to 15% for nursery exit weights; and

8 to 12% for weight at first pull from the finishing barn.
In other words, if the CV for body weight at weaning is around 20%, at nursery exit is 1215%, or at first pull is 812%, then it makes sense for barn managers to seek ways to manage variability rather than reduce it because it is already close to the lower practical limit.
These targets may change in the future as we develop more information on CVs under commercial conditions.
Topic of Interest
Variability is becoming a topic of increased interest in the pig industry due to its substantive impact on net income. This has become more obvious in allin, allout systems, where tailend pigs are more obvious than they would be in a continuousflow system.
Although variability has earned extra attention, there is much to learn, such as how much variability is inherent and must therefore be accepted, and how much is excessive and therefore, at least theoretically, can be reduced.
Poor herd health, itself a poorly defined term, is believed to be a major contributor to variability. But we also know that inadequate access to feed and water can contribute.
Managing variability will be the focus of most farms, requiring imaginative strategies based on facts and not specious logic to be successful.
Unfortunately, many procedures known to effectively manage variation require either changes in physical facilities or increased labor, neither of which is viewed with much enthusiasm at this point in the pork production cycle.
Nonetheless, there are substantial rewards to producers who reduce or manage variability effectively, because the costs of not doing so are also substantial.
Average age, days  

19  68  140  
No. of pigs  1,264  700  632M  
Weight, lb.  
Mean  11.9  64.1  228.2  
Median  11.9  64.0  229.7  
Minimum  5.3  52.4  163.7  
Maximum  20.2  90.0  274.8  
Range  14.9  37.6  111.1  
Range, % of mean  121  59  48  
Standard deviation, lb.  2.7  8.2  18.3  
Coefficient of variation, %  22.4  12.82  8.02  
1Body weights were determined on whole groups of animals without preselection at weaning (19 days), nursery exit (68 days of age) and at 20 weeks of age before the first market pull. All were collected at the Prairie Swine Centre Elstow Research Farm. The weights were collected at different times, so the three groups of pigs are not related to each other. The ages represent means, although the range in ages would be plus/minus three or four days. 