Scribd
Upload
18 views

Types of Statistical Tests by Purpose

Uploaded by

DR
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
18 views

Types of Statistical Tests by Purpose

Uploaded by

DR
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 4

.

Types of Statistical Tests by Purpose

 Descriptive Tests: Summarize data (mean, median, standard deviation).


 Comparative Tests: Compare means, medians, or proportions between groups.
 Correlation and Association Tests: Measure the strength and direction of relationships.
 Regression Tests: Predict outcomes and assess relationships.
 Nonparametric Tests: Analyze data without assuming a specific distribution.
 Advanced and Specialized Tests: Multivariate tests and tests for complex designs.

2. Overview of Key Tests and When to Use Them

A. Descriptive Statistics

 Mean, Median, Mode: Central tendency measures; used to describe the average values in a dataset.
 Variance and Standard Deviation: Measures of variability; useful for understanding data spread.

B. Comparative Tests

These tests are used to compare means, proportions, or distributions between two or more groups.

1. t-tests:
o Independent t-test: Compares the means of two independent groups (e.g., male vs. female patient
temperatures). Use when data is normally distributed and variances are equal.
o Paired t-test: Compares means of two related groups (e.g., before-and-after measurements for the
same patients).
o One-sample t-test: Tests if the mean of a single group is different from a known value (e.g.,
comparing patient temperature mean to a standard).
2. ANOVA (Analysis of Variance):
o One-way ANOVA: Compares means across three or more independent groups (e.g., knowledge
levels across multiple education groups). Assumes normality and equal variances.
o Two-way ANOVA: Assesses the effect of two different categorical independent variables on one
continuous dependent variable (e.g., comparing the effects of knowledge and experience on patient
outcomes).
o Repeated Measures ANOVA: Like a paired t-test but for more than two time points or conditions.
3. Chi-Square Test:
o Chi-Square Test for Independence: Used for categorical data to test if two variables are
independent (e.g., is there an association between gender and knowledge level).
o Chi-Square Goodness of Fit Test: Tests if a categorical variable’s observed frequencies match
expected frequencies.
4. Mann-Whitney U Test:
o Used to compare medians between two independent groups when data is not normally distributed
(nonparametric alternative to the independent t-test).
5. Wilcoxon Signed-Rank Test:
o Nonparametric alternative to the paired t-test, comparing two related samples.
6. Kruskal-Wallis Test:
o Nonparametric alternative to one-way ANOVA, comparing medians among three or more
independent groups.

C. Correlation and Association Tests

These tests evaluate relationships or associations between variables.

1. Pearson Correlation:
o Measures linear correlation between two continuous variables (e.g., age and temperature); assumes
normality.
2. Spearman’s Rank Correlation:
o Nonparametric correlation that assesses monotonic relationships between two continuous or ordinal
variables.
3. Kendall’s Tau:
o Nonparametric test measuring strength of association between two ordinal variables, especially with
small sample sizes.
4. Chi-Square Test of Association:
o For categorical variables, evaluates if there’s an association between them (e.g., association between
intervention type and patient outcome).

D. Regression Tests

Regression tests predict outcomes and quantify relationships between variables.


1. Linear Regression:
o Models the relationship between one dependent continuous variable and one or more independent
continuous/categorical variables. Useful when predicting a continuous outcome, such as patient
recovery time based on age and intervention type.
2. Logistic Regression:
o Used for binary outcomes (e.g., success/failure); predicts the probability of an outcome based on
independent variables.
3. Multiple Linear Regression:
o Extends linear regression to multiple independent variables, helpful for adjusting for confounders.
4. Cox Proportional Hazards Regression:
o Common in survival analysis; models the time to an event based on predictor variables (e.g., survival
time of patients based on intervention type and age).

E. Nonparametric Tests

These are ideal when data doesn’t meet normality assumptions.

1. Sign Test:
o Compares median of a sample to a known value or compares paired observations without requiring
normality.
2. Friedman Test:
o Nonparametric alternative to repeated measures ANOVA for comparing multiple related groups.

F. Advanced and Specialized Tests

These tests handle complex data structures.

1. Multivariate Analysis of Variance (MANOVA):


o Extends ANOVA to multiple dependent variables, useful when assessing the effect of one or more
factors on several outcomes simultaneously.
2. Mixed-Effects Models:
o Useful for hierarchical or clustered data (e.g., repeated measures or data with random effects like
patient groups). They handle data where responses are not independent.
3. Survival Analysis (Kaplan-Meier and Cox Regression):
o For time-to-event data, Kaplan-Meier estimates survival function, while Cox regression accounts for
covariates affecting survival time.

How to Choose a Test

1. Type of Variable(s):
o Continuous, categorical, or ordinal variables impact the choice. For example, continuous outcomes
often use t-tests, ANOVA, or regression.
2. Number of Groups:
o Comparing two groups? Use t-tests or Mann-Whitney U. More than two groups? Consider ANOVA
or Kruskal-Wallis.
3. Normality of Data:
o Normally distributed data allows for parametric tests; if not, use nonparametric alternatives.
4. Paired vs. Independent Samples:
o Paired samples (e.g., pre- and post-tests for the same group) use paired tests like the paired t-test or
Wilcoxon signed-rank, whereas independent samples use independent t-tests or Mann-Whitney U.
5. Study Design:
o Observational studies may benefit from correlation or regression to adjust for confounders, while
experimental studies can use causal tests like t-tests, ANOVA, or logistic regression.

You might also like