Sampling Techniques in Veterinary Medicine

The process of obtaining data in veterinary investigations is referred to

as measurement.

What is a Census:

• In a census, every animal in the population is evaluated.

• E.g., all dogs in a shelter are assessed for prevalence of ectoparasites.

In a census, the only source of error can be what

the measurement itself

A census can be very what

expensive and time consuming.

In a sample data are collected from what

only a subset of the population

Collecting data from a subset of the population is more convenient

than collecting data from the entire population.

What kind of errors can we get from a sample

With a sample, you have both measurement and sampling error to contend with.

Samples are drawn from a population to support both what kinds of investigations

descriptive and analytic investigations

What is A descriptive investigation:

describes population attributes e.g., prevalence of a disease in a canine shelter, or cow herd.

What is the defintion of Analytic investigation:

estimates the magnitude of the association between outcomes and exposure factors in a population.

In veterinary investigations, there are 3 populations of interest, What are they

• targe population

• source population

• study sample/group

What do we mean by Target population:

• It is the population to which it might be possible to extrapolate results from a study/investigation.

The target population is often not well what ?

well-defined & may vary depending on who is interpreting the results of the study.

What is source population:

• It is the population from which the study subjects are drawn.

What do we mean by Study sample or group:

• This consists of individuals (animals or groups of animals) that end up in the

study.

• This group is usually a sample from the source population.

Prior to the investigation, an appropriate sample size is

determined and individuals in this group are selected using an appropriate sampling method.

What do we mean by Internal validity of an investigation:

• Relates to whether or not the investigation results obtained from the selected sample are valid for members of the source population

• It indicates whether or not the study has obtained the correct answer for the

source population.

What do we mean by External validity of the investigation:

• Relates to how well those results can be generalized to the target population.

Assessing external validity asks the question:???

Are the results representative of the target population?

A sampling frame is a list of

all the sampling units in the source population.

Sampling units are what

the basic elements of the population that is sampled e.g., herds, animals.

A complete list of all sampling units is required in order to do what

• draw a simple random sample,

• but may not be necessary for some other sampling strategies.

The sampling frame is the information about the

source population that enables you to draw a sample.

The objectives of the investigation will influence the

sampling strategy employed.

You should always have SMART objectives WHAT DOES S mean

Specific

You should always have SMART objectives WHAT DOES M mean

Measurable

You should always have SMART objectives WHAT DOES A mean

Achievable

You should always have SMART objectives WHAT DOES R mean

Relevant

You should always have SMART objectives WHAT DOES T mean

Time-bound

In investigations based on a sample of observations, the results we obtained are affected by:

• The variability of the outcome being measured.

• Measurement error

• Sample-to-sample variability.

There are two types of error: what are they

• Type I (alpha) error

• Type II (beta) error

What are Type I (alpha) errors:

You conclude that the outcomes in the groups being compared are different (i.e an association exists) when in fact they are not.

What are Type II (alpha) errors:

You conclude that the outcomes are not different (i.e no association between the outcome and the exposure exists) when in fact they are.

Statistical tests aim at disproving the what kind of hypothesis

null hypothesis

Null hypothesis: there is no difference between groups, If differences are found, they are reported with a

p-value

P-value is the probability of what

that a difference as large or larger than the one observed could be due to chance if the null hypothesis is true.

P-value is the probability of making a what kind of type of error

type I (alpha) error.

The alpha level can be set at

0.1, 0.05, 0.01.

When P-value is less than the set alpha level, e.g., P < 0.05, we are reasonably sure that the

observed effect is not due to chance.

What do we mean by Power

The power of a study is the probability that you will find a statistically significant difference when it exists and is of a certain magnitude.

Power = what ? Which type of Type of error

• 1-beta (1-type II (beta) error

• Power is usually set at 80%.

Reasons why a study may fail to detect a difference between groups

• There truly was no effect of the exposure on the outcome

• The study design was inappropriate

• The sample size was too small (low power)

• Chance

What do we mean by Non-probability sampling:

are drawn without an explicit method for determining an individual’s probability of selection.

A non-probability sample is one that is drawn without what

a formal process for random selection.

Non-probability samples are not appropriate for what

descriptive studies except for initial pilot studies.

Non-probability sampling procedures are often used in what kind of studies

analytic studies.

There are 3 types of non-probability sampling: what are they

• Judgement sample

• Convenience sample

• Purposive sample

What do we mean by Judgement sample:

• Chosen because in the judgement of the investigator, it is representative of the source population.

• The criteria for inclusion and the selection process is not explicit.

A convenience sample: is what

• Chosen because it is easy to obtain study subjects

• Herds with good handling facilities

• Herds with easily accessible records.

A convenience sample is used with what kind of studies

• Used in analytic studies where the need to have a representative sample of the source population can be relaxed.

What is the definition of Purposive sample:

• The selection of this type of sample is based on the study subjects possessing one or more attributes such as known exposure to a risk factor or a specific disease status.

• Often used in observational studies.

What is Probability sampling

Probability sampling is where every element in the population has a known non-zero probability of being included in the sample.

Different types of probability sampling include: what

• Simple random sampling

• Systematic sampling (a.k.a systematic random sampling)

• Stratified random sampling

• Cluster sampling

• Multi-stage sampling

What is the definition of Simple random sampling

In simple random sampling, every study subject in the source population has an equal chance of being included.

A formal random process is used to select subjects & these are:

• Drawing numbers from a hat

• Using computer generated random numbers e.g. in MS Excel

• Flipping a coin

• Throwing a dice

• Using a random numbers

What do we mean by Systematic sampling

• In systematic sampling, a complete list of the source population is not required

• All you need is an estimate of the total number of animals, & all animals or their

  records

You then calculate the sampling interval (j) as:

Study population size divided by required sample size.

What is Stratified random sampling

In stratified random sampling, the population is divided into mutually exclusive strata based on factors likely to affect the outcome.

In its simplest form, the number sampled within each stratum is proportional to

the total number in the stratum-this is called proportional sampling

What are the Advantages of stratified random sampling

• Ensures that all strata are represented in the sample

• The precision of the estimates might be greater than those derived from a simple random sample

• It produces estimates of stratum-specific outcomes.

What is the defintion of Cluster sampling

A cluster is a natural or convenient collection of study subjects with one or more characteristics in common e.g.,

In cluster sampling, the primary sampling unit is

larger than the unit of concern.

In cluster sampling, every study subject within the cluster is included or excluded in the sample

• included in the sample.

• Individuals within a cluster are more likely to be alike

Cluster sampling is done because it is easier to get what

get a list of clusters (shelters, farms) than to get a list of individuals.

It is less expensive to sample a larger or smaller cluster

smaller number of clusters than to collect information from many different clusters.

When the group is both the sampling unit & the unit of concern, then that

sample is consider what

not a cluster sample e.g., a sample of herds to determine herd prevalence of a disease is not a cluster sample.

What is Multi-stage sampling

Multi-stage sampling is similar to cluster sampling except that after the primary sampling unit is chosen, a sample of secondary sampling units is selected.

In other words, in cluster sampling, is ?

• First, clusters are selected

• Second, sampling units within the selected clusters are randomly selected.

What do we mean by Targeted (risk-based) sampling

Targeted (risk-based) sampling is used in animal disease surveillance programs especially for rare diseases

Targeted sampling involves stratification of the source population into what

strata based on one or more characteristics which are thought to be associated with the probability of disease occurrence.

Only strata in which the probability of finding what

finding cases of a disease is highest are sampled.

Targeted sampling are extensively used in what animal

BSE (bovine spongiform encephalopathy) surveillance programs.

What do we mean by Sample-size determination

• The choice of sample size involves both statistical & non statistical considerations.

Non-statistical considerations include: what

• the availability of resources such as time, money, sampling frames

• Some considerations of the objectives of the study

• Costs can be factored into sample-size calculations

• When the budget is fixed, & the cost per study subject is high, the sample size will be small.

Statistical considerations include what

• The required precision of the estimate. The more precise you wish to be, the larger the sample size.

• The expected variation in the data (the variance expected in the outcome of interest)

• The desired level of confidence

What is the formula for Estimating proportions or mean

What is the formula for comparing proportions or means

What do we see in Precision-based sample size computations

• Here, sample size calculation is done by specifying the desired width of the confidence interval for the difference being estimated.

What do we see in Power-based sample size computations

• Here, sample size is calculated by specifying the desired power of the study to detect a difference of a defined magnitude.

What do we mean by Sampling to detect disease

Sampling to detect the presence or confirm the absence of a disease is fundamentally different than sampling to estimate a parameter e.g., prevalence of a disease.

If a finite population is being sampled, the formula is:

If sampling from a large or infinite population, the formula is:

At an alpha = 0.05, you say that you are

95% confident that the prevalence of the disease in the population is below the minimum threshold you specified for this disease.

Errors may lead to what

invalid results & effects can range from minor to profound.

Errors are classified as:

1. Random error

2. Systematic error

What is the definition of Random error:

Divergence, due to chance alone, of an observation on a sample

from the true population value, leading to lack of precision in the measurement of an association

What are the sources of random error

Sampling error, biological variation & measurement error

What are examples of Systematic error (bias):

Selection bias, information bias & confounding

Understanding sources of error is important WHYYY?

• So, you don’t get fooled by WRONG information from research studies or your clinical/epidemiologic “investigations”

What is the defintion of Bias

• Systematic error in a study that results in an incorrect estimate of the association between a risk factor/exposure of interest and the health outcome/disease being studied.

What are examples of bias

1. selection bias

2. information bias

What is the definition of Selection Bias

results from how study subjects are selected for inclusion in the study

What is the definition of Information Bias

results from inaccuracies in obtaining information/data on study parameters

What is confunding

• The mixing of the effects of one risk factor/exposure with another

Confunding results in what

• Results from an extraneous factor that has an association with the exposure and outcome of interest