EDS6003 Chap.9 Descriptive and Inferential Analysis
Descriptive and Inferential Analysis
Description before inference
The survey report asks students to describe participants and results, then use inferential reasoning such as correlation, a t-test or an F-test where appropriate. Counts, distributions, means and variation show what was observed in the achieved sample.
Inferential analysis asks how compatible a relationship or difference is with chance variation under a model, but it does not erase weaknesses in sampling or measurement.
Let the question select the statistic
Correlation describes a linear relationship between variables. A t-test compares two means relative to within-group variability. An F-test can assess differences among several group means.
Statistical significance does not report effect size, educational importance or causation. Interpretation should therefore begin with the observed pattern and then state what the design permits.
What this chapter covers
- 01
Counts and distributions
- 02
Mean and median
- 03
Correlation and association
- 04
Two-group t-test
- 05
Several-group F-test
- 06
Statistical significance and causal limits
Worked example · free
Compute a mean and inspect its distribution
- 3Sum the values and divide by the count.
- 2Find the median and compare it with the mean.
- 3State the bounded interpretation.
Key terms
- Mean
- The arithmetic balance point found by dividing the sum of observations by their count.
- Median
- The middle ordered observation, which is less affected than the mean by an extreme value.
- Correlation
- A measure of the direction and strength of a linear relationship between two variables.
- T-test
- An inferential method comparing two means relative to variation within the groups.
- F-test
- An inferential method comparing variation among several group means with variation within groups.
Descriptive and Inferential Analysis FAQ
When should a mean be reported?
A mean is useful for a quantitative measure when its scale and distribution make arithmetic averaging meaningful. Report sample size and spread, inspect skew and outliers, and consider the median. Decimal precision cannot compensate for a poorly measured construct.
What does a statistically significant result mean?
It means the observed result would be comparatively unusual under the specified null model and assumptions. It does not state that the effect is large, educationally important, measured well or caused by the predictor. Those questions require separate evidence and interpretation.
How do I choose between correlation, a t-test and an F-test?
Use correlation for a relationship between measured variables, a t-test for a mean difference between two groups, and an F-test when comparing several group means. Begin from the research question and variable structure, then inspect assumptions rather than selecting the output with the smallest probability.
What belongs beside a statistical test result?
State the achieved sample, compared groups or variables, units, descriptive pattern and observed magnitude before the probability statement. Add uncertainty and the main design limit in the same account. A small p-value does not repair selective sampling, weak measurement or a non-causal design. The reader should be able to reconstruct what was compared and why the chosen procedure answers the research question.
Assessment move
Calculate small examples by hand before opening statistical software. For each output, write the research question, variables, observed pattern, method and one-sentence interpretation. Add a second sentence stating what the result cannot establish. This habit prevents significance from replacing substantive reasoning and makes numerical errors easier to detect. Build an analysis card for each research question.
List the observation unit, variable type, groups or paired structure, expected display and candidate statistic. Before calculating, sketch the distribution you would need to see for the proposed summary to be informative. Then create one awkward alternative containing skew, an outlier or a curved relationship and explain how the same statistic could mislead.
Recompute a mean from five values and verify the sum, count and units. For a correlation or group comparison, describe the visible pattern before naming the test. Keep selection, measurement and causal limits in a separate box so they remain present when a probability value appears. Practise reporting a result in three layers: observed magnitude, uncertainty under the chosen model and educational meaning.
None can substitute for the others. When several tests appear plausible, return to the wording of the question and the data-generating design rather than choosing from software menus. End with a reconstruction check: another student should be able to identify the compared groups, outcome scale, achieved sample and direction of the finding from your results sentence alone.
Working through Descriptive and Inferential Analysis in EDS6003? Sia is AskSia’s AI Education tutor — ask any EDS6003 Descriptive and Inferential Analysis question and get a clear, step-by-step explanation grounded in how EDS6003 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
