Research Methods and Statistics in CSD

Comprehensive practice flashcards covering parametric and nonparametric inferential statistics, descriptive statistics, qualitative research designs, correlational and developmental methods, survey and replication designs, and evidence-based practice evaluation.

What is the primary distinction between descriptive statistics and inferential statistics?

Descriptive statistics summarize and measure features of a specific dataset, whereas inferential statistics infer characteristics of an entire population based on a sample subset.

What are the three main categories of descriptive statistics?

1. Measures of location (mean, median, mode) as a single value
2. Measures of individual location (ranks, standard scores)
3. Measures of variability (scatter plots, data dispersion around a central location)

What foundational sampling assumption underlies inferential statistical procedures, and how can researchers address its violation?

Samples are presumed to be randomly selected, meaning every member of the population has an equal chance of being chosen. If random sampling is not possible, researchers can resolve limitations by carefully matching groups on key extraneous variables.

Which statistical tests represent the most common parametric inferential statistics used in CSD research, and what level of data measurement do they require?

Student's $t$-tests and Analysis of Variance (ANOVA). Both require continuous data measured at least at the interval or ratio level.

Why are parametric statistics generally preferred over nonparametric statistics, and when do nonparametric tests offer MORE statistical power?

Parametric tests rely on stronger continuous data and normal distributions, providing greater statistical power and more reliable outcomes when data are near-normal. Nonparametric tests provide more statistical power when the underlying sample distributions are highly skewed.

What type of statistical test is appropriate for categorical (nominal) count data, and why is it considered 'distribution-free'?

The Chi-square test ($\chi^2$). It is distribution-free because it does not rely on the assumption that data are sampled from a normally distributed population.

What are the key assumptions of the Chi-Square test?

1. Observations must be independent nominal counts
2. Categories must be mutually exclusive
3. Expected cell frequencies must not be too small (typically $E \ge 5$)
4. The sum of expected frequencies must equal the sum of observed frequencies

Based on the contingency table format shown below, how are observed and expected frequencies presented, and what formula calculates the cell deviation?

The first number in each pair is the Observed count ($O$) and the second number is the Expected count ($E$). The cell deviation is calculated as: $\frac{O - E}{E}$

What specific research design requires a $t$-test for related (paired) samples versus a $t$-test for unrelated (independent) samples?

A related samples $t$-test evaluates a single group tested across two conditions (e.g., pre-test/post-test design). An unrelated samples $t$-test evaluates a two-group (between-subjects) design comparing means across two separate, non-overlapping groups.

What are the nonparametric alternatives to the related samples $t$-test and the unrelated samples $t$-test?

The Wilcoxon Test for Paired Data is the alternative for related samples, and the Mann-Whitney U test is the alternative for two unrelated samples.

What general sample size threshold indicates using the Student's $t$-statistic rather than the $Z$-statistic or ANOVA?

Sample sizes less than 30 ($n < 30$) indicate the use of the Student's $t$-distribution, whereas sample sizes greater than 30 ($n > 30$) use the $Z$-statistic or ANOVA.

What type of distribution curve is depicted below, which describes a Student's $t$-distribution that is narrower and taller than the standard normal curve?

A leptokurtic curve.

How is the $t$-ratio mathematically calculated in a Student's $t$-test for unrelated samples, and what does the denominator represent?

The $t$-ratio is calculated by dividing the difference between the means of the two groups by the standard error of the difference. The standard error of the difference represents the variability of scores due to error, determined by the pooled variance of the two groups.

What is the conventional cutoff alpha level for statistical significance in CSD parametric research, and what does homogeneity of variance require?

The conventional cutoff is $p < 0.05$ (5 percent confidence level). Homogeneity of variance requires that the two comparison groups have equal or near-equal amounts of variability (variance) around their means.

What corrective procedure can researchers use if continuous data violate the assumption of a symmetrical, normal distribution?

Apply data transformations (mathematical operations) to achieve a more normal distribution.

When comparing exactly two group means, what is the mathematical relationship between the ANOVA $F$-statistic and the $t$-statistic?

$$F = t^2$$

How is the ANOVA $F$-statistic mathematically calculated as a ratio of variance components?

$$F = \frac{\text{Between-Group Variability}}{\text{Within-Group Variability}} = \frac{\text{Mean Square Effect}}{\text{Mean Square Error}}$$

In a simple one-way ANOVA, what is the difference between 'within-group variability' and 'between-group variability'?

Within-group variability is the portion of total variance that cannot be explained by the research design (mean square for error / variability within each group). Between-group variability is the portion of variance attributable to group membership or the experimental treatment (mean square for effect).

What is the primary nonparametric alternative to a simple one-way ANOVA, and under what conditions is it especially appropriate?

The Kruskal-Wallis Test (yielding the $H$-statistic). It is used to compare three or more independent groups when sample sizes are small, unbalanced, or uneven.

What is a complex (multifactor) ANOVA, and how do 'main effects' differ from 'interaction effects'?

A complex ANOVA evaluates two or more independent variables, each containing two or more levels. Main effects evaluate each independent variable independently, whereas interaction effects evaluate how combinations of variables influence one another.

Why are post hoc comparison tests necessary following a significant omnibus $F$-test in ANOVA, and what are four common examples?

A significant $F$-test indicates that at least one significant difference between group means exists, but does not specify which specific pairs differ. Four common post hoc tests are: Bonferroni, Tukey's, Newman-Keuls, and Scheffé.

Why are the Scheffé and Bonferroni post hoc tests frequently criticized?

They are overly conservative, effectively testing closer to $\alpha = 0.01$ even when nominally set at $\alpha = 0.05$.

What is the nonparametric alternative to a complex (multifactor) ANOVA, and what is its main limitation?

Friedman's Test (which ranks ordinal data for related groups). Its main limitation is that it cannot evaluate interaction effects.

How does a 'simple effect size' differ from a 'standardized effect size'?

A simple effect size is the raw difference between treatment means, which ignores variability and rarely permits cross-study comparisons. A standardized effect size divides the mean difference by the pooled variance, allowing direct comparisons across different studies.

Which standardized effect size statistics are recommended for large sample sizes versus small sample sizes?

Cohen's $d$ is used for large groups ($n > 30$), and Hedges' $g$ is used for small samples ($n < 30$).

What is meta-analysis, and why is it frequently conducted in CSD?

Meta-analysis is a quantitative method that aggregates standardized effect sizes across multiple studies to calculate an overall average effect size, commonly used to examine treatment effectiveness across the literature.

What core methodological purpose differentiates qualitative research from quantitative research?

Qualitative research focuses on generating theories, describing experiences, and exploring phenomena in naturalistic settings, whereas quantitative research focuses on testing pre-formulated hypotheses.

What is 'inductive analysis' in qualitative inquiry?

An analytical process that begins with specific observed examples and facts, progresses through open-ended questions, and concludes with the development of general theories.

What does 'thick description' mean in qualitative research methodology?

Gathering detailed, highly context-rich information about the phenomenon of interest and triangulating data from multiple sources.

What is 'empathic neutrality' on the part of a qualitative researcher?

Positioning oneself as a neutral, non-judgmental observer who acknowledges personal biases, remains aware of them, and tempers their influence during interpretation.

How does 'lamination' differ from basic 'triangulation' in qualitative data verification?

Triangulation compares data collected at different times, methods, or locations. Lamination goes one step further by cross-referencing findings through data collection, forming tentative conclusions, and then asking participants their thoughts about the interaction.

What defines Ethnography as a qualitative research design?

A qualitative approach that focuses on the entire culture of a group of people, typically requiring the researcher to become immersed as an active participant observer.

What is the defining premise of Phenomenology?

A philosophical qualitative design that rejects speculation about unobservable phenomena and focuses exclusively on phenomena as immediately lived and experienced by human subjects.

What is Grounded Theory?

A problem-solving qualitative approach that iteratively collects, verifies, and analyzes observational data until a credible, explanatory theory is developed directly from the data.

What is Discourse Analysis?

A qualitative design concerned with language use beyond the boundaries of an individual utterance or sentence, such as conversational interaction strategies.

In kinesic analysis, what are emblems, illustrators, and adaptors?

Emblems are nonverbal gestures with specific, direct meanings. Illustrators are body movements that accompany and time the rhythm of speech. Adaptors are unconscious body movements performed under specific circumstances (such as tucking hair behind an ear when anxious).

How does the Direct Observation Method differ from the Participant Observation Method?

Direct Observation involves systematic observation of participants without researcher involvement in the scene (e.g., through one-way mirrors). Participant Observation requires the researcher to actively participate and become an accepted member in the culture or setting.

What is 'referential adequacy' in establishing the credibility of qualitative research?

Setting aside a portion of unanalyzed data until tentative conclusions are drawn, then checking whether the conclusions match the unanalyzed data.

What are 'member checks' in qualitative research?

Sharing final findings, conclusions, and interpretations directly with study participants to evaluate how closely they align with the participants' lived reality.

What defines 'transferability' in qualitative studies?

The extent to which qualitative study results can be applied or transferred to other persons, settings, locations, events, or contexts.

What are the four models for combining qualitative and quantitative methods in multi-method research?

1. Sequential: qualitative methods followed by structured quantitative methods
2. Parallel: qualitative case studies conducted alongside quantitative methods
3. Coordinated sub-studies (CSS): various qualitative studies coordinated as part of a larger program project
4. Integrated: qualitative and quantitative data and concepts integrated at every stage

What three properties do correlation coefficients describe regarding the relationship between variables?

1. Strength (magnitude or amplitude, ranging from $0$ to $1.00$)
2. Direction (positive or negative)
3. Statistical significance (meaningful or non-significant)

In CSD correlational research, what correlation coefficient thresholds denote meaningful relationships?

A correlation coefficient of $0.30$ may be considered meaningful, and a correlation of $0.70$ or higher is always meaningful.

When is the Pearson product-moment correlation ($r$) used versus the Spearman rank correlation ($r_s$)?

Pearson's $r$ is used for continuous data measured on interval or ratio scales with normal distributions. Spearman's $r_s$ is used when data are measured on ordinal scales (such as ranks or ratings).

What does the coefficient of determination ($r^2$) represent?

The arithmetic square of the correlation coefficient, representing the proportion of variance in one measure attributable to or shared with the other measure.

What are the roles of the 'criterion variable' and the 'predictor variable' in path analysis and multiple regression?

The criterion variable is the outcome being predicted, while the predictor variable is the variable used to make the prediction.

What is the primary objective of Factor Analysis?

To analyze a large set of correlated variables in order to identify fewer, underlying common factors.

What is the typical independent variable in developmental research designs, and how is it operationally defined?

The independent variable is maturation, operationally defined by chronological age, reflecting the underlying physical and mental changes that occur over time.

What is the primary threat to internal validity in longitudinal developmental designs versus cross-sectional designs?

In longitudinal designs, the main threat is participant attrition over the extended study duration. In cross-sectional designs, the main threat is group nonequivalence across different age cohorts.

How does a semi-longitudinal developmental research design compromise between cross-sectional and longitudinal approaches?

It includes two or more groups representing different points along the developmental continuum (like cross-sectional) but follows each group across a shorter multi-month duration, typically $3$ to $12\text{ months}$.

What is the difference between a sample survey and a census?

A sample survey collects data from a representative sample of a population, whereas a census surveys each and every member of the population.

How does stratified sampling reduce sampling error in survey research?

By dividing the population into logical strata or sub-divisions to ensure that each group is proportionally and fairly represented in the final sample.

What are the three main types of replication designs and their primary purposes?

1. Direct replication: duplicates a study in every detail using new participants to establish the reliability of previous results
2. Systematic replication: duplicates a previous study with purposeful variations (e.g., settings, participant populations) to establish generality
3. Conceptual (derived) replication: tests the same theoretical concept or hypothesis using different methods to establish theoretical validity

What is 'Information Literacy' as defined by the Association of College and Research Libraries?

A set of abilities requiring individuals to recognize when information is needed and have the ability to locate, evaluate, and use the needed information effectively.

In Evidence-Based Practice (EBP), what differentiates 'raw evidence' from 'pre-filtered evidence'?

Raw evidence is information that has not been subjected to expert review, whereas pre-filtered evidence has been critically evaluated by domain experts (such as peer-reviewed journal articles).

What are the three standard subsections typically contained within the Method section of a quantitative research report?

1. Description of participants
2. Description of apparatus/equipment and materials/stimuli
3. Description of procedure (step-by-step summary of execution and manipulations)

In clinical treatment research, how does 'efficacy' differ from 'effectiveness'?

Efficacy evaluates how well a treatment works under ideal, highly controlled laboratory conditions. Effectiveness evaluates how well a treatment works in routine, real-world clinical practice settings.

What does a confidence interval indicate in a research report?

It measures the precision of a statistical estimate, indicating the range within which the true population parameter is expected to lie.

When comparing outcome score distributions, what does 'less overlap' between two distributions indicate?

Less overlap indicates greater variation and a larger difference between the two measures, groups, or treatment conditions.