Module 1

Question 1

What four methods does Neuman (2011) outline that people use to aquire knowledge?

Answer 1

1. Personal Experience and common sense: can be distorted through perception, illusion and judgement
   
   • Common sources of bias: overgeneralisation, premature closure, the halo effect, false consensus
2. From Experts and Authorities: Easily accept information from authority figures as true
3. Peers and Media Messages: Distorted by level of education of source of information, bias in reporting and cultural myth
4. Ideological Beliefs and Values: views shaped by beliefs are often presented as factual knowledge

Question 2

How does Neuman define the scientific method?

Answer 2

• Science is both a system and the knowledge derived from that system.
  
  • Theory = coherent system of ideas used to condense knowledge
  • Data = form of empirical knowledge collected via science
  • Empirical = evidence/observations grounded in sensory data
  • Scientific Literacy = capacity to understand and use scientific method
  • Innumeracy = lack of quantitative literacy
• A scientific approach is characterised by:
  
  • Clearly defined constructs
  • Theory driven research Qs
  • Falsifiable hypotheses
  • Valid meausurement of constructs
  • Reproducible design, methods and results
  • Peer review and consensus

Question 3

What are the five basic norms of the scientific community according to Neuman?

Answer 3

• Norms are informal rules/principles/values that govern the way scientists conduct research
  
  • Universalism: all research should be judged on merit regardless of researcher or institution
  • Organised Skepticism: challenge and question all evidence
  • Disinterestedness: neutrality, impartialiality, openness and receptiveness to unexpected/new ideas
  • Communalism: Knowledge belongs to all, it should be shared and critiqued by all and peer reviewed
  • Honesty: no dishonesty or cheating in research

Question 4

What biases and non-scientific sources of information should be avoided according to Neuman?

Answer 4

• Overgeneralisation: Statement that goes far beyond what can be justified based on the data
• Selective Observation: Examination that reinforces preexisting thinking
• Premature Closure: Making a judgment or ending an investigation before
  gathering sufficient evidence
• Halo Effect: When prior experience colours one’s evaluations rather than evaluating all in a neutral, equal manner
• False Consensus: assuming that everyone else thinks like he or she does
• Pseudoscience: Information with outward appearance of science that was not created with systematic rigor
• Junk Science: A public relations term used to criticize scientific research that produces findings that an advocacy group opposes.

Question 5

What are the differences between quantitative and qualitative research?

Answer 5

Question 6

What are some common research designs?

Answer 6

• Non-experimental: Does not involve manipulation of an independant or dependant variable
• Correlational: Examining variables through interpretation of their association (no causality).
  
  • Examine Direction, Strength and Statistical Significance (all interrelated)
• Quasi-experimental: Use pre-existing or other non-randomly assigned groups or interventions. ie a correlational study with manipulation
• Experimental: allow inference of causal relationships between variables. The exprimenter actively manipulates the independent variable(s).

Question 7

What is the difference between exploratory, descriptive and explanatory research?

Answer 7

• Exploratory: Primary purpose is to examine a little understood area and develop preliminary ideas.
  
  • Addresses “what” questions
  • Formulate and focus questions for future research
  • Most uses qualitative data (surveys, interviews)
• Descriptive: Presents a picture of the specific details of a situation, social setting, or relationship.
  
  • Focus on “how” and “who” questions (ie how often, describe patterns)
  • Use most data-gathering techniques: surveys, field research, content analysis, and historical-comparative research
• Explanatory/Experimental: To explain why events occur and to build,
  elaborate, extend, or test theory.
  
  • Testing a (usually causal) hypothesis

Question 8

What are five potential errors in causal explanation?

Answer 8

• Tautological errors: the causal factor and the result are actually the same or restatements of one another
  
  • eg Sally is conservative because she believes in less regulation
• Teleological errors: the causal relationship is empirically untestable because
  the causal factor occurs after the result or cannot be empirically measured
  
  • Crime occurs because it is just human nature.
• Ecological Fallacies: When data at a higher or an aggregated unit of analysis isd used to explain something about a smaller or disaggregated unit
  
  • Country A has more motorbikes than B; therefore Person from A does.
• Reductionistic errors: when a person observes a lower or disaggregated
  unit of analysis but makes statements about higher units
  
  • eg Civil Rights Movement due to Martin Luther King Jnr alone
• Spuriousness: occurs when two variables are associated but are not causally related because an unseen third factor is the real cause
  
  • eg IQ causes TV vs Reading and Level of Knowledge

Question 9

What is a hypothesis? What are the different types of hypothesis?

Answer 9

• A hypothesis is an empirically testable statement predicting the relationship between 2 or more variables
  
  • Research determines support never proof. Research is a dynamic process of generation, investigation and elimination of hypotheses.
  • 5 factors of a Causal Hypothesis: 2+ variables, causal relationship, predictive, linked to theory and falsifiable.
• Types of Hypotheses
  
  • Null v Alternative: Ho predicts no relationship
  • Directional V Non Directional: based on previous research either direction or not (increased X –> decreased Y)
  • Double Barrel Hypothesis: Poorly designed. Two IVs mean 3 predictions.
    
    • Use an Interaction hypothesis instead

Question 10

What are the different types of missing data?

Answer 10

• Missing Completely At Random (MCAR): no pattern in the missing data
• Missing at Random: a pattern to missingness not associated with the DV
• Missing Not at Random: a pattern associated with the DV
• How to find out which in SPSS:
  
  • Analyse -> Missing Value Analysis -> Add experimental group to categorical -> Estimation EM (chi-square)
  • Look for the MCAR test under EM Means -> if p-value is significant it is MCAR

Question 11

How can you deal with Missing Data?

Answer 11

• List-Wise Deletion: deleting the case (not good option, deletes all other data)
• Mean: Replace value with the mean of the item
• Common-Point: Replace value with the midpoint of the scale
  
  • Transform -> Recode into Same Variable -> Old and New Variables -> Select System or User Missing and add the midpoint value
• Multiple Imputation (best option): Using regression to predict the response based on their other data and the group
  
  • Analyse -> Multiple Imputation -> Impute Missing Data Values
  • Options: Add all IVs and DVs into model, 1 imputation, New Data Set
  • Constraints: min and max values of scale, use as predictor and impute and use as predictor

Question 12

What are the four possible reasons for outliers?

Answer 12

1. Data Entry Errors (always screen for data entry errors before anything else)
2. Not Specifying missing-value codes in SPSS (always screen missing values first)
3. The outlier is not a member of the population you intended to sample
4. The outlier is from the intended population but the distribution for the variable is more extreme than a normal distribution

Question 13

What are the different types of outliers and how can you deal with them?

Answer 13

• Univariate Outlier: A response on DV > 3.29 SDs from the mean of the group
  
  • Split cases: data -> select cases -> if experimental group = 0
  • Analyse -> descriptive stats -> DVs -> save standardised
• Multivariate outlier: responses across DVs indicate inconsistent responding
  
  • Analyse -> Regression Linear -> Reverse IV’s and DVs -> Save MAHs distance
  • MAH -> Transform -> Compute Var -> prob = 1- CDF.CHISQ(MAH_1, #predictors)
    
    • Look for probability <0.001
• Dealing with Outliers
  
  1. ​Check for errors in data entry
  2. Fix univariate outliers; Winsorizing (replace with normal score)
  3. Remove Multivariate outliers altogether.

Question 14

What is the assumption of Normality?

Answer 14

• Normality assumes that the random error (residuals) are normally distributed throughout the data set (not the data itself)
  
  • Especially important when n<200 (before Central limit theorem sets in)
  • Affects Type 1 Error (false positive)
• Distributional information
  
  • Skewness: symmetry of distribution (positive skew = clustered at low end)
  • Kurtosis: peakedness of the distribution (leptokurtic = tall,slender) (platykurtic = flat)

Question 15

How can you test for Normality?

Answer 15

• It is reccomended to use multiple methods of normality
• Shapiro Wilk test: Ho is a normal distribution. If the result is significant, the error distribution is not normal
  
  • Split the groups: Data -> Select case -> if condition satisfied
  • Analyse -> Explore -> Group filter -> plots -> normality plots
• Multivariate normality: no direct test in SPSS
  
  • If each DV is normal via Shapiro Wilk, generally overall is normal
  • Can run a multivariate normality test via Syntax if needed

Question 16

What is Homoscedasticity/homogeneity of varience?

Answer 16

• Both refer to equal variance across predictors, homoscedascity is used for regression designs and homogeneity of variance for group effects
  
  • Violation results in biased confidence intervals and significance testing
• Assessment
  
  • Zpred v Zresid plot: scatter plot of the z-scores of errors against z-scores of predicted values. Triangle shape indicates heterogeneity
  • Levene’s Test: Compares Significance. Non significant result indicates homogeneity
  • Fmax Ratio: Ratio of largest variance to smallest variance compared to a critical value

Question 17

What is linearity, additivity, independence and multicollinearity?

Answer 17

• Linearity: the relationship is ebst described by a straight line
• Additivity: Combined effects of predictors is best described by adding them
• Independence: Errors in a model are not related
  
  • Affects confidence intervals and significance testing
  • Observations should be independent
• Multicollinearity: When 2+ variables are very highly correlated (if they are a perfect fit it is singularity)
  
  • Cutoffs vary between r > 0.7 to r > 0.9

Question 18

What are some methods to reduce bias according to Riddiford?

Answer 18

• Trimming Data: detection of cases with extreme values (either by % or SD)
  
  • New Mean = the Trimmed Mean. Only use when outlier is non representative
• Winsoring: Extreme scores replaced by normal score. Several methods and definitions exist
• Robust Methods: not affected by violations of assumptions
  
  • Non-parametric Statistics: don’t rely on normal distributions
  • Robust Methods: Bootstrapping (estimate parameters)
• Transformation: Alters value of each score to address assumptions
  
  • Different options for each issue
  • Issues: incorrect transformations, change hypothesis, small sample sizes

Question 19

What are non-parametric statistics?

Answer 19

• Parametric Stats rely on
  
  • Large enough data set (min 20-30 per group)
  • Homogeneity of varience
  • The data are normally distributed
  • Cells are of equal size
• Nonparametric stats
  
  • are based on ranks
  • should be used with ordinal data or when sample size is small
  • required sample size is dictated by the analysis