Research Methods Reference

Regression analysis estimates how independent variables affect a dependent variable using a fitted equation. It comes in simple, multiple, logistic, and hierarchical forms depending on the number of predictors and the type of outcome, and it proceeds by checking assumptions, fitting the model, and interpreting coefficients.

Correlation analysis expresses the direction and strength of the linear relationship between two variables as a correlation coefficient (r) ranging from -1 to +1. Choose Pearson or Spearman based on the measurement scale and normality, and remember that correlation means association, not causation.

A t-test checks whether the difference between two group means, or between one group mean and a reference value, is statistically significant. It comes in one-sample, independent samples, and paired samples forms. Check normality and equality of variances first, then judge the result from the t statistic and its p-value.

Analysis of variance (ANOVA) tests mean differences across three or more groups in a single test, and is classified by the number of factors into one-way and two-way designs. If the F-test is significant, post hoc tests identify which specific groups differ.

Factor analysis uncovers the small number of latent factors behind many measured items. It splits into exploratory factor analysis (EFA), which discovers the structure, and confirmatory factor analysis (CFA), which tests a theoretically specified structure. The procedure runs from suitability checks through extraction, rotation, and naming.

Structural equation modeling (SEM) is a multivariate technique that tests causal relationships among latent variables by splitting the analysis into a measurement model and a structural model. It accounts for measurement error and lets you test complex theoretical models, including mediation and moderation, in a single estimation.

The chi-square test uses the difference between observed and expected frequencies to test association or distributional fit for categorical variables, and divides into goodness-of-fit and independence tests. When expected frequencies are small, Fisher's exact test is used instead.

Time series analysis examines trend, seasonality, and autocorrelation in data observed in time order to forecast future values. It proceeds through checking stationarity, identifying the model, estimating it, and forecasting and diagnosing, with ARIMA as the representative model.

Cronbach's alpha is an internal consistency reliability coefficient showing whether several items measure one concept consistently, and 0.7 or above is generally treated as acceptable. A low value calls for reviewing item deletion, while an excessively high value suggests redundant items.

Cluster analysis is an exploratory method that groups cases with similar characteristics into a few homogeneous clusters, divided mainly into hierarchical and k-means clustering. It proceeds by standardizing variables, computing distances, forming clusters, and deciding the number of clusters.

Difference-in-differences (DiD) estimates a causal effect by comparing the before-after change in a treated group with the change in an untreated control group. It is a quasi-experimental design that rests on the parallel trends assumption, and it proceeds by defining groups and timing, checking pre-trends, estimating an interaction-term regression, and running robustness checks.

Meta-analysis statistically combines the results of multiple prior studies on the same topic to estimate an overall effect size. Studies are gathered through a systematic review, after which effect sizes are computed, pooled, and tested for heterogeneity and publication bias.

The Delphi method builds consensus by surveying a panel of experts anonymously over several rounds, feeding results back each time so opinions converge. It is used where prediction is difficult or where criteria and indicators must be developed, and consensus is judged with measures such as the content validity ratio (CVR) and convergence.

SPSS is statistical software that runs everything from descriptive statistics to t-tests, ANOVA, regression, and factor analysis on survey and experimental data through menu clicks. It is the most widely used analysis tool for papers in the social sciences and health fields.

AMOS is software that analyzes structural equation models by drawing them on screen as diagrams. It is widely used for confirmatory factor analysis, path analysis, and testing mediation and moderation, and it reports a full range of model fit indices.

The PROCESS macro is an extension for SPSS and R that tests mediation, moderation, and moderated mediation through bootstrapping. You specify the analysis type with one of Hayes's model numbers and judge significance from the confidence interval of the indirect effect.