Data Collection Methods

Observational research: gathering information through observation, usually used to answer a question on a topic where little is known, answer a how or a why question

  • Qualitative: interview, observations

  • Quantitative: collecting numerical data

Surveys

  • Likert scale: answering from ‘strongly disagree’ to ‘strongly agree’ - each assigned a numerical value

  • A scale score is obtained when answers to all of the survey questions are combined - eg IQ score. IQ score is a constructed variable, as it was constructed from all the other answers.

Sampling

Non-random sampling

External validity of non-random samples: low

  • Convenience sampling: participants are chosen based on ease and convenience

  • Purposive sample: researchers look for individuals with specific characteristics and carry out case studies on them

  • Quota sample: a convenience sample with a condition for numbers within groups

  • Snowball sample: participants suggest other participants to the researcher

  • Stratified sample: selecting a sample of different subgroups in the population that represents the proportions of those groups in society

Random sample: participants are selected at random, sample is representative for the entire population. High external validity

  • Simple random sample: a computer is used to randomly select participants. Drawback: a list of everyone who could possibly be a participant is needed

    • Coverage errors: when a part of the population is not covered by the list used during the sampling process

    • Non response errors: participants refuse to answer certain questions or the whole survey. May yield biased results

  • Cluster sample: people are divided into arbitrary groups, groups are selected and all participants within that group are tested

  • Systematic sample: a random starting point is found and people are selected with a fixed, periodic interval

  • Stratified random sampling: strata are purposefully selected, and then individuals are randomly chosen from the strata

  • Multi-stage sampling: similar to cluster sampling

  • Oversampling: subgroups within the population are purposefully overrepresented, and then the data is adjusted to fit the proportion of that subgroup to the rest of the population in real life

Measurement methods

Example: researching people’s eating habits

Survey

  • Problems: people forget

Observations in a lab

  • Hawthorne effect: people behave differently in the lab than normal

  • Only represents a snapshot in time

  • Problems with external validity

Taking photos of groceries

  • Lots of context and information, however low reliability

Keeping a food diary

On paper

  • Non response

  • Low reliability - people don’t often tell the truth and people often forget what they ate

Online/through an app

  • Coverage error - not everyone has a phone/can use it

  • Nonresponse error