It would be ideal to score every animal on the farm. However, the ideal becomes impractical on large herds, unless we endeavored to spend several days on a farm. Using a targeted sampling method, we can score a subset of animals to generate a prevalence estimate. We determine the subset to sample based on the risk for each concern.

## Goals:

- To accurately estimate the prevalence of a problem in the herd
- To target the population that is at risk for that particular problem

We face some constraints when we subsample. There are key concepts that are important to know:

**Score by exception**. For example, in the case of locomotion it is possible to score every cow in a given group as they exit the parlor by exception. That is, observing every cow, but only recording the number of cows with a score of 2 or 3. Because the total number of cows in the pen can be known, we can assume any cow NOT counted as a 2 or 3 is a 1.**Score when animals are restrained**(only applies to measures other than locomotion). By scoring when animals are restrained, they are easier to count, and often easier to assess. Forms of restraint will depend on the age class and farm, but they include tie stalls, routine use of headlocks and milking parlors.**Determine minimum number within the target age class**. When animals are unrestrained, it can be difficult to score each one. To address this, a sample size calculator can be used to determine the minimum number of animals to evaluate.**Try to eliminate bias**. When scoring groups of unrestrained cattle, it is important to be aware of several types of bias.*Human attention is often drawn to extremes:*animals with no problems or those with obvious ones. Having a rule of thumb can help avoid this problem. For example, scoring every 2nd animal you encounter.*Sometimes animal behavior introduces bias*. For example, lame cattle are more likely to be lying down than sound ones, so if only standing animals are evaluated, the group estimate will appear better than it actually is. Another example is that severely lame cattle are more likely to be in the last third in the milking order, so inclusion/exclusion of this portion of a milking string affects the group estimate. Being aware of these types of patterns will help you minimize bias.

## Groups to Score

In the table below, ideal samples should be taken if time and the housing system allows; minimum sampling is required regardless of time taken.

When calculating farm prevalence for each measure and life-stage group, use the total number of animals of that group observed as the denominator.

When calculating farm prevalence for each measure and life-stage group, use the total number of animals of that group observed as the denominator.

Additional measures not included in the above chart, including dehydration, do not have a required sample size. Instead, any animals in hospital pens will be assessed to ensure that they have adequate access to food and water.

## Determining Sample Group

On farms with 100 or fewer animals in a life-stage (milk-fed calves, heifers, or lactating cows), all the animals in that life-stage will be scored. On farms where there are more than 100 animals in a life-stage, we will apply a select group sampling approach.

Using the sample size calculator, round the group number UP to the nearest value in the table.

## Example

Here are some examples about how to calculate sample sizes: Farm A has fewer than 100 animals per life-stage while Farm B has over 100 animals per life-stage.