Chapter 4- Sampling Issues

Snowball Samples

- used to study groups or individuals that are hard to identify or locate - a drawback, members often known each other so there is potential bias in the sample

Cluster sampling

- used when it is difficult or impossible to list the members of the target population

When something is truly random

- you get clusters of things because its weird, its what it is - when your sample is small and the phenomenon is random you might land on a blank spot - when your sample gets bigger your confidence interval gets smaller because you are sure

Why sampling?

-Of those who are invited to participate, only some fraction agree to do so, and are able to complete the sudy. -That group is the sample. It is inevitably different from the sampling frame in some ways.

Randomness

-The fundamental nature of randomness is that it cannot be explained, that there is no pattern to predict. • Randomness, therefore, does NOT lead to evenness (sameness); it leads to unpredictable variety. It will produce some extreme values. • Therefore, a bigger sample size can overcome some of these random variations

Population parameter

-actual value of the mean, standard deviation, and so on, for the population -usually population parameters are not known, though often they are estimated based on information in sample

Convenience Samples

-composed of participants who are easily available - the problem is that it's never clear to what population the findings apply

Response Bias

-people from the sampling frame who agree to participate are different from those who decline • The response rate (percentage of those invited who agree to participate) is a clue to how much response bias will be involved - The people that agree are just like us or they are comfortable with the person - people who respond are different then the ones that don't

Purposive Samples

-requires a judgment or an "educated guess" as to who should represent the population - heavily dependent on a comprehensive understanding of the processes under study by the researcher

Volunteer Bias

-the concept that people who volunteer to participate in research studies often differ from those who do not volunteer - tend to be optimistic, try new things

Sample statistic

-the empirical value determined for the elements included in our sample

Sampling error/ bias

How the scientist determines the sampling frame, and how they recruit the participants, both have a big impact on the results • Because each introduces the possibility of sampling error (AKA sampling bias), which is any way that the subset is different from, and not generalizable to, the population as a whole. • Sometimes this matters a lot. Sometimes it matters less.

How to Evaluate a Sample?

Is it big enough? • Due to randomness, bigger samples give us more reliable estimates. 2. Is it representative of the population of interest? Or is there a sampling bias? - Note: due to the ethical requirement that participation by voluntary, there is almost always a response bias.

Sampling Strategies

Probability • All members of the sampling frame have an equal likelihood of being invited to participate • Essential if the research is attempting to determine prevalence of something. • Non-Probability • Good enough for research attempting to explore relationships between variables.

Is a bigger sample always better?

Size helps shrink confidence intervals around numerical estimates • This is critical when estimating prevalence • But size does not reduce bias • Therefore, response rate is usually a more important issue than sample size.

Selection Bias

something about the way the participants were selected introduces a systematic difference between participants and the population • Case in point: the sampling frame is skewed • E.g., bias toward white, middle-class women in family science • E.g., college sophomore problem - how is the sampling frame different from the whole

Example

- Average Family Size Population- people in fresno Sampling Size- students in CFS153, convenience which is non-bias Sampling Bias- Hispanics have larger families, mothers that don't have children

Nonprobability Sample

- No way to estimate sampling error

Probability Sampling (every member of the sampling frame has the same probability of being selected)

- Simple Random Sampling - Every single member of population has equal chance of being selected. 2. Stratified Random Sampling - Population is divided into groups. Every member of a group has equal chance of being selected. -Necessary when a subgroup of interest is very small. 3. Cluster Sampling - Every cluster has equal chance of being selected, every person in a selected cluster has equal chance of being selected. - Necessary when the population is very large.

Determination of Sample Size

- amount of sampling error that can be tolerated - amount of variability in the dependent variable - cast of gathering data about additional observations

Sample Size

- as it INCREASES, the amount of ERROR Decreases, which is GOOD - but Cost also INCREASES which is BAD

Bias

- bunch of biases built into a sample - every study has layers of biases, some are easy to notices others are not

Population

- complete and inclusive collection of all theoretically defined elements - the group of interest, target you care about - usually impossible and unnecessary to study everyone in the population

Sampling error

- differences between the characteristics of the sample and those of the population

Sample is

- different from the frame and the frame is different from the population

Probability samples

- error generally decreases with increasing sample size, but cost increases with sample size

Simple random sampling

- every element in the sampling frame has a known and equal probability of being included in the sample

Probability sampling

- every member of the target population has a known and nonzero probability of being included in the sample, selection must be random

College Sophomore Problem

- example of a bias - they are often participants in psychology research, but they dont represent the whole population

Sampling Procedures

- help us decide who is studied and how to select those individuals to produce valid results that can be generalized to the entire population under study

Quota Samples

- involves creating a sample whose characteristics mirror those of some population of interest - can be constructed to oversample groups of interest

Nonprobability sampling

- likelihood of any given element being included in the sample is unknown, and the selection process is NOT necessarily random

Units

- refer to the analysis of the data

If a study has

- response bias you can look at the rate of response, which is the percentage who agreed to participate

Sampling Bias

- sampling frame - sample

Why narrow down

- since you cant study the whole population you create a sampling frame to narrow it down - we assume the sampling frame is changing the study, but we just don't know

Sampling procedures

- techniques through which we determine which persons or families to observe and how many to observe

Sample

- the collection of elements drawn from the population that are actually studied - consists of the people who agree and participate in the study

Disproportionate stratified random sampling

- the researcher can obtain a greater proportion

Sampling Frame

- the set of all elements from which the members of the sample, actually studied are drawn - goal to obtain a sample whose data will yield results similar to those that would have been obtained if data were collected on the entire population of interest - narrowing down the population, a smaller group, a limit

Element

- the unit about which information is gathered - linked to the sample- selection process

Stratified random sampling

-the population is divided up into groups, called strata, then a simple random sample is drawn from each stratum - based on characteristics such as zip code, country, or gender

4 major types of Nonprobability Sampling Procedures

1. Convenience Sampling 2. Quota Sampling 3. Purposive Sampling 4. Snowball Sampling

Non-Probability Sampling

Convenience Sampling -Study those who are easy to access 2. Purposive Sampling Select from identified groups 3. Snowball Sampling - Get a participant, ask him/her to suggest another -Used when a sampling frame is very hard to identify 4. Quota Sampling -Match proportions in the population, but use convenience sampling to get those numbers