Validitet and Reliabilitet

• Validitet is how well a study measures what it intends to measure
• Reliabilitet is how reliable and stable the measurements are

Intern vs. Ekstern Validitet

• Intern validitet is whether the study establishes a causal relationship between variables
• High internal validity means control over confounding factors, ensuring changes in the dependent variable are due to the independent variable
• Ekstern validitet is whether the study's results can be generalized to other situations, groups, or contexts
• High external validity means results apply beyond the specific study to a larger population or different settings
• There is a trade-off between intern and ekstern validitet

Ensuring High Reliabilitet in a Study

• Reliabilitet ensures consistent measurements that yield the same results upon repetition
• Achieve high reliabilitet by defining procedures, using standardized measurement tools, and avoiding subjective assessments
• Quantitive studies strengthen reliabilitet by using test-retest methods, inter-rater reliabilitet (comparing multiple observers' assessments), and statistical methods like Cronbach's alpha to test internal consistency
• Qualitative studies strengthen reliabilitet through documentation of the research process, data triangulation (using multiple sources), and preventing researcher bias

Limitations of Various Research Methods

• Eksperimentelle studier often have high internal validitet but low external validitet due to artificial settings
• Tverrsnittstudier can show correlations but not causation, weakening internal validitet
• Panelstudier provide information on changes over time but may suffer from high attrition, affecting both validitet and reliabilitet
• Qualitative methods provide deep understanding, but often have low reliabilitet because findings are context-dependent and harder to replicate
• Spørreundersøkelser may have low validitet if questions are unclear or if respondents give socially desirable answers
• Researchers minimize limitations by conducting methodological evaluations and ensuring the research design fits the study's purpose

Problemstilling

• A problemstilling is a precise question that research aims to answer
• It narrows the topic and makes it researchable
• It should be clear, concrete, and formulated as a question
• A good problemstilling indicates what is being examined, who is being studied, and is narrowly defined

Hypotese

• A hypotese is a testable assumption about the relationship between two or more variables
• Hypotheser are used in quantitative studies to test causal relationships
• They should be precise and verifiable
• A good hypotese should be testable with data, indicate an expected direction (positive/negative correlation), and be concrete
• A problemstilling is a question to be answered through research
• A hypotese is an expected relationship that can be tested empirically

Difference between Problemstilling and Hypotese

• Problemstilling: "What is the relationship between social status during upbringing and perceived merkepress?"
• Hypotese: "The higher the social status during upbringing, the higher the perceived merkepress."

Importance of Problemstilling and Hypotese

• A problemstilling provides direction for the research
• A hypotese allows testing and drawing scientific conclusions
• Be able to formulate a problemstilling, explain the difference between a problemstilling and hypotese, and create a hypotese from a topic.
• Målenivåer determine how variables are measured and which statistical analyses can be used
• Nominalnivå involves categorical variables where values represent different groups without ranking
  • There is no logical sequence between values
• Ordinalnivå involves variables with a natural ranking, but the distance between values is not necessarily equal
  • One value is higher or lower than another, but not by how much
• Intervallnivå involves variables where the distance between values is equal across the scale, but there is no absolute zero point
  • Differences between values are comparable, but one value isn't twice as large as another
• Målenivået influences the choice of analysis method and meaningful calculations
• Nominalnivåvariabler are often analyzed with frequency tables
• Ordinal- and intervallnivåvariabler can be used in more advanced analyses like averages and regression models

Ethical and Legal Considerations

• Research involving humans must follow ethical and juridiske guidelines to protect participant rights and safety
• Informert samtykke includes providing participants with sufficient information about the study, including purpose, method, risks, and potential consequences, before they agree to participate
• Participants must also be informed they can withdraw at any time without consequences
• Personvern focuses on protecting participant identity and personal data
• Data must be stored securely, and researchers must follow laws regulating data collection, storage, and use
• Forskningsetiske prinsipper comprise respect for participant autonomy, harm prevention, fairness, and scientific integrity
• Research must be relevant, minimizing risk to participants
• Researchers must be objective and avoid conflicts of interest
• Research violating ethical or juridiske guidelines results in severe consequences, such as invalid results, loss of trust, and juridiske sanctions

Identifying Causal Relationships

• Kausalitet is about establishing a cause-and-effect relationship between variables Data analysis is used to examine such relationships and interpret results based on statistical methods
• To establish that variable X affects variable Y, three criteria must be met:
• Samvariasjon – X and Y must be correlated
• Tidsrekkefølge – X must precede Y to be a potential cause
• Utelukkelse av alternative forklaringer – Other factors that could affect Y must be controlled
• Eksperimentelle designs provide the strongest basis for establishing kausalitet because they isolate influencing factors
• In observational studies, control variables and statistical analyses must be used to reduce the risk of misinterpretation

Regresjonsanalyse, Korrelasjon, and Signifikansnivå

• Korrelasjon measures the degree of correlation between two variables, but doesn't necessarily establish kausalitet
• A high korrelasjon may be due to chance or an underlying variable
• Regresjonsanalyse is used to examine how one or more independent variables affect a dependent variable, helping to control for other factors
• Signifikansnivå (often set at 5%) is used to determine whether an observed relationship is statistically significant, i.e., the probability that the result is due to chance is low
• Regresjonsanalyser provide information about the strength and direction of the relationship between variables, as well as whether the findings are statistically significant

Målenivåer: Nominal, Ordinal, Intervall

• Nominalnivå - Variables with different categories without ranking, where the values only represent groups (e.g., residence or profession)
• Ordinalnivå - Variables with categories that can be ranked, but where the distance between categories is not equal (e.g., educational level)
• Intervallnivå - Variables where the distance between values is equal, but there is no absolute zero point (e.g., temperature measured in Celsius)
• Målenivået determines which analyses can be used
• Nominal- and ordinaldata are often analyzed with frequency tables and cross-tables
• Intervallnivåvariables can be analyzed with averages, variance analyses, and regression models
• Research design is the structure for how a study will be conducted to answer a research question
• The design determines how data is collected, analyzed, and interpreted

Main Types of Research Designs

Research designs are largely divided into kvalitative and kvantitative designs.

1. Kvalitative research designs (exploratory, deep understanding) These designs are used when in-depth knowledge about a phenomenon is desired Interviews, observations, or text analysis are often used

• Fenomenologi - Studies experiences and learnings from individuals' perspectives
• Case-studie - In-depth analysis of one or a few cases (e.g., an organization, a person, a society)
• Etnografi - Studies culture and social practices through fieldwork and observation
• Narativ forskning - Analyzes stories and tales from informants
  • Benefits: Deep insight, flexibility
  • Drawbacks: Time-consuming, difficult to generalize

2. Kvantitative research designs (measurement, numbers, and hypotesetesting) These are used when aims to quantify data and analyze causality

• Eksperimentelt design- Controls and manipulates variables to study causality and effect.
• Ekte eksperiment: Randomized control group (e.g., medical studies)
• Kvasieksperiment: No full randomization, but still controlled setting
• Tverrsnittstudie – Collects data at one point in time, often via questioning
• Longitudinell studie – Collects data at multiple points in time to study changes over time
• Panelstudie: Same persons are followed over time
• Gjentatte tverrsnitt: New selections are drawn at each measurement
• Korrelasjonsstudie – Examines relationships between variables (e.g., is there a connection between training and life quality?)
• Benefits: Objectivity, generalizability, possibility for statistical analysis
• Drawbacks: Can lack depth of understanding, risk of misinterpreting causality

Kvantitative and Kvalitative Methods

1. Kvantitative methods

• Kvantitative methods are used when aiming to collect measurable data and examine relationships between variables.
• Examples of kvantitative methods:
• Spørreundersøkelser: Standardized questions with fixed answer options, e.g., opinion polls or surveys about health habits.
• Eksperimenter: A group receives an impact (e.g., a new teaching plan), while another doesn't, to see if it has an effect.
• Registerdata: Use of existing statistics, such as public data about income or education.
• Benefits: Many respondents, generalizable findings.
• Drawbacks: Limited depth, can overlook nuances.

2. Kvalitative methods

• Kvalitative methods are used when aiming for in-depth knowledge and understanding of how people think and feel about something
• Examples of kvalitative methods:
• Intervjuer: In-depth conversations where participants tell about their experiences, e.g., about bullying in school.
• Observasjon: The researcher observers and notes how people behave in a situation, e.g., in a classroom.
• Dokumentanalyse: Analysis of texts, e.g., public documents or media articles.
• Benefits: Gives rich insight and understanding.
• Drawbacks: Time-consuming, difficult to generalize.

Sannsynlighetsutvalg vs. Strategisk Utvalg

Sannsynlighetsutvalg (Random Selection)

• Used in kvantitative studies to ensure the sample mirrors the population
• Every unit has a known probability to get selected
• Examples:
• Enkel tilfeldig trekking: All individuals in the population have equal chance of getting selected (e.g., random selection of 1000 people from the population register)
• Stratifisert utvalg: The population is divided into groups (e.g., age, gender), and random selection happens from each group
• Klyngeutvalg: Entire groups are selected randomly, e.g., randomly selected schools in a study about the school environment
• Benefits: Gives generalizable results
• Drawbacks: Can be difficult to implement if the population is large or hard to reach

Strategiskutvalg (Targeted Selection)

• Used in kvalitative studies to select informants who can provide relevant insight
• Examples:
• Maksimal variasjon: Select people with diverse backgrounds to get breadth in the answers (e.g., young and old job seekers)
• Typiske eller ekstreme case: Select units that are particularly interesting (e.g., a top manager or a person who has experienced extreme poverty)
• Snøballmetoden: One participant recommends new participants (e.g., interviews with people in a subculture)
• Benefits: Gives rich and detailed data
• Drawbacks: Not generalizable to the entire population

Securing Representativeness in the Sample

Representativeness means that the sample mirrors the population, to use the results to say something about the whole group.

• Measures:
• Use probability selection, so that everybody has an equal chance of getting selected
• Have a large enough selection to avoid random skjevheter
• Avoid systematic skjevheter, so that only one group answers (e.g., if only older people participate in a study of technology.)
• Do frafallsanalyse to check if those who did not respond differ from those who answered
• Example of bad representativeness:A study about political attitudes where only university students participate

Summary

• Statistical methods provide breadth and measurable data, while qualitative methods give deep insight
• Probability selection is used for generalizable studies, while strategic selection is used for deeper understanding
• To secure statistical methods, you must avoid skjevheter in who participates and how the selection is drawn