SDA: Studies and Sampling techniques

Question 1

Cross-Sectional Studies

Answer 1

Observe phenomenon at a particular point in time

Whole population, or sample of population can be observed

Question 2

Longitudinal/Cohort Studies

Answer 2

Sample taken and monitored over time
Prospective design: future
Retrospective design: past

Question 3

Samples should be:

Answer 3

Large and representative enough to make INFERENCES and CONCLUSIONS that are GENERALISABLE to a whole population

Question 4

Why do we sample?

Answer 4

Can't count/enumerate a whole population due to:
Resources
Time
Cost
Logistics

Question 5

Is the census a sample?

Answer 5

Some argue it represents a ‘kind of sample’ of reality
Undercount and hard to reach population subgroups
Once every 10 years - not representative? Miss key changes?

Question 6

When do we use statistical testing?

Answer 6

To draw conclusions for:
A Single Sample: e.g. is a class mean test score statistically significant?
Paired Observations: see if there is a change in outcome between two measuring occasions
Two Samples: e.g. is there a significant difference between house prices for two locations?
More than two samples: e.g. does grocery expenditure significantly vary between several locations?

Question 7

Inferential/Statistical testing

Answer 7

Statistical inferences made about a whole population based on sample results

Take a sample of whole population then take sample statistics to make statistical estimates of population parameters

Question 8

When making inferences, we calculate the chances of being wrong in what ways?

Answer 8

Sceptical decision making surrounding research hypotheses
Risk of chance findings
Are results statistically significant?
Ability to accept research hypotheses
Is there enough information/evidence to draw conclusions/make inferences

Question 9

Population parameters

Answer 9

Population mean, median, variance etc.

Question 10

Sample statistics

Answer 10

Sample mean, median, variance etc.

Question 11

What should a good sample achieve?

Answer 11

Data that provides estimates that are precise and on target
Decrease bias
Sample estimates to be close to population values
Best sampling involves some element of random selection

Question 12

Probabilistic Sampling

Answer 12

Known chance that an individual will be included in the sample
Two most common types: random and stratified
Stratified sampling usually carried out initially - i.e. the population is divided into subsets and then either systematic or random sampling occurs in each subset

Question 13

Simple Random Point Sample

Answer 13

Randomly select x and y coordinates
Uses point location
PROBLEM: could end up with spatial clustering, uneven coverage and be imprecise

Question 14

Systematic Point Sample

Answer 14

Randomly starting point (grid reference appearance)
Subsequent points selected at regular, pre-determined distances e.g. every 4 points
PROBLEM: may miss variability

Question 15

Arealy Stratified Point Sample

Answer 15

Combination of random and systematic
Population is divided into cells/strata
Randomly sample within the cells
GOOD: more evenly sampled, distributed and precise

Question 16

2-Stage Area Cluster Point Sample

Answer 16

Divided into cells/strata and randomly sample cells before randomly sampling points within these
Used in large-scale social surveys
PROBLEM: could miss local specificity, give imprecise and inaccurate results

Question 17

Stratified, Systematic, Unaligned Point Sample

Answer 17

Combination of all sample designs

Method:

1. Divide data into cells/strata
2. Randomly selected number gives x-coordinate and is kept constant across the first row of grid squares, y-coordinate varies in each strata and selected at random
3. First column, the randomly selected y-coordinate is kept constant and the x-coordinate varies
4. Method then repeated for every row and column until a point location has been selected in each strata

Most accurate and precise estimates of population
Method is precise if there is a good areal coverage

Question 18

How do you choose which sampling technique to use?

Answer 18

As long as the sample is representative of the data, the sample estimates will be accurate
If there is a strong spatial auto-correlation (i.e. patterning) then the techniques will give varying results, therefore need to carefully consider technique taking into account: cost, time, resources, logistics