Quanti data collection

Five Steps in Data Collection Process

• Determine Participants
• Secure permissions from individuals/organizations
• Consider types of information to collect from sources
• Locate and select useful instruments
• Administer the data collection process

Identifying Participants

• Identify the unit of analysis, consider: individual, family, school, school district.
• Specify the population and sample

Population, Target, and Sample

• Population: The group of individuals sharing a characteristic.
• Target Population/Sampling Frame: The list of sampling units from which a sample is selected.
• Sample: The group of study participants selected from the target population

Probability Sampling

• Involves random selection of a sample from a population’s members.
• Aims to examine representative elements of populations.

Simple Random Sampling

• A probability sampling method that ensures equal and independent chance of each population element being chosen.
• Ways to select a simple random sample: Fishbowl, table of random numbers, calculator.

Fishbowl Technique/Lottery Method steps

• Write numbers for all participants on same-sized, colored papers.
• Roll papers and place in a container.
• Shake container thoroughly and draw a number.
• Cease when the desired sample number is picked.

Fishbowl Technique Procedures

• Random sample with replacement: drawn paper with the number is returned.
• Random sample without replacement: drawn paper with the number is not returned.

Table of Random Numbers Method

• Uses a list of numbers generated without order or sequencing.

Table of Random Numbers Method Steps

• Assign number to each element of accessible population.
• Enter table at a random starting point.
• Continue systematically (up, down, left, right, diagonally) until sample is selected.
• Select numbers in a chosen direction up to the desired number is reached.
• Skip numbers if encountered more than once.

Calculator use to generate random numbers

• Use NRan#
• N stands for population size.
• Ran# is the key for generating random numbers.

Systematic Random Sampling

• Involves selecting every kth element of the population.
• K = N/n, where K is the interval, N is the population size, and n is the sample size.

Systematic Random Sampling Steps

• Construct population frame; assign each participant a number.
• Compute the interval K using K = N/n.
• Enumerate elements in the N set and include a number for random start.
• Select a number using the Table of Random Numbers.
• Select every nth population member using the number as system
• Cease when the required participant number is obtained.

Stratified Random Sampling

• Population divided into subgroups (strata) related to research study variables.
• Steps: construct population, determine strata, select number using stratified sampling, choose category participants with simple random methods.

Two sample approaches from strata

• Proportional Stratified: sample from each stratum is proportional to its size in population.
• Disproportional Stratified: the number of members chosen is not in proportion to size of the total population.

Multistage Cluster Sampling

• Cluster Random Sampling, large groups or "clusters" become sampling units.
• Example order from largest area to smallest: Region, Province, Town, Barangay

Power Analysis

• Correctly rejects a false null hypothesis in a statistical test's probability.
• The probability a test identifies treatment effect if one exists.

Factors that Influence Power

• Effect Size, as effect increases, Ho is rejected.
• Sample Size, a larger sample produces higher test power.
• Primary reasons for computing power is to determine sample size to achieve reasonable probability for study.
• Alpha Level, reducing this will reduce power.
• One-Tailed vs. Two-Tailed Tests, effects increase power by switching from regular two-tailed test to one-tailed if in the predicted direction.

G*Power Analysis

• Used to compute statistical power analyses for t tests, F tests, x2 tests, z tests, and exact tests.
• Computes effect sizes and displays power analysis results graphically.
• Main purpose is to determine sample size needed to detect the given test effect at the desired significance level.

G*Power example with three pieces of information

• Conduct a cross-sectional study to measure the correlation between overtime hours worked and burnout symptoms in emergency room nurses.
• Sample size estimates to determine correlation significance from zero.
• Three pieces of information required: estimate of the correlation size (r), a two-tailed alpha value (a), and a beta value (β)

Two-tailed alpha value (a)

• Type I error rate: the risk of having non-zero correlation when the effect is not real
• Alpha value is typically .05 or a 5% probability

Beta value (β)

• Type II error rate-the risk of stating no significant effect when it exists
• Beta value is typically .20
• Estimated power asks for 1 – beta ( or 1 - .20 = .80)

Plugging in values from Power calculator

• Can determine the sample size needed. By using the calculator to determine the sample size with r = .25, with a two-tailed alpha value of .05 and a beta value of .20
• Power analysis will determine the need for at least 123 participants.
• Use G*Power using the software program with following input parameters
  • Test family: Exact
  • Statistical test: Correlation: Bivariate normal model
  • Type of power analysis: A priori: Compute required sample size
  • Tails: Two
  • Correlation p Η1: .25
  • a err prob: .05
  • Power (1 – ẞ err prob): .8
  • Correlation p H0: 0

Nonprobability Sampling

• Sample elements are chosen by nonrandom methods.
• Individuals are representative of the population.
  • Convenience: readily available people/objects.
  • Snowball (network/chain): participants identify others to be in a sample.

Nonprobability Sampling continued again

• Quota: Similar to stratified random sampling.
• Purposive: aka judgmental sampling, involves "handpicking" of subjects,.

Sample Size Considerations in Quantitative Studies

• Larger sample is most representative.
• Larger sample is a smaller sampling error.
• As sample increases, probability of getting odd samples diminishes (sampling bias).
• Power analysis can determine the sample size.

Sample error

• The difference between: a random sample of data vs If an entire population were measured

Sampling Bias

• Occurs when samples are not carefully selected.

Sample Size Estimates (Creswell, 2019)

• 15 participants per group in an experiment.
• 30 participants for a variable correlational study.
• 350 individuals for a study that uses a survey, but the size may vary.

Sampling Designs in Qualitative Studies

• Convenience and Snowball: uses volunteer or nominated sampling.
• Purposive: Select the sample purposefully.
• Strategies of Purposive Sampling Patton 2015): Maximum, Homogenous, Typical case, Extreme/Deviant Case, Confirming and Disconfirming Cases.
• Theroretical: most used in grounded theory

Data saturation

• the guiding principle in qualitative sampling.

Sample Size in Qualitative studies

• A small amount an analysis, or analysis
• Final decisions during collections
• studies: 25-50 informants, phenomenology: 10, Grounded : 20-30

Sample Description in Qualitative studies

• Should include sampling procedure, population under eligiblity, studies and rationale

Types of Permissions required

• Permission from institutions
• Specific sites.
• Participants, parents, or campus

Review Board Approval

• An committee instititional of faculty
• Obtaining Approvals
• Star by process from boards
• Determine needs for and of of form

Types of Qualitative Data and Measures

• Measures Performance, types of test. intelligence etc...
• Measures, observations, or and school

Selecting the appropriate instrument

• Steps developing: phases objective, evaluate contruct
• Chossing an 5 widely endorse document test

Instrument Modification

• To locate, and use
• Request of cite

Reliability and Validity of an Instrument

• Stable scores, unambiguous, standardized test.
• Validity: Evidence test matches use

V. The process

• Types 2 nominal scales