The University of Manchester · FACULTY OF ACCOUNTING

BMAN30071 Chap.2 Regression Tools for Market Evidence

- one subject, every graph, every model, every mark
6 Chapters2-page Bible
Our own words - no uploaded lecturer files
Updated for this semester
Chapter 2 of 10 · BMAN30071

Regression Tools for Market Evidence

Regression is the module's main empirical language. This chapter explains fitted values and residuals, separates coefficient magnitude from statistical precision, interprets explanatory power, and reconstructs group-specific slopes from dummy-variable interactions. It also traces how measurement error in different variables changes fit or biases the relation being estimated.

A complete table reading starts with the dependent variable, units and base group before moving to stars or test statistics. Interaction terms measure differences from a base relation, so subgroup effects require explicit addition. Economic size, uncertainty and overall fit answer separate questions.

Measurement error also has variable-specific consequences: noise in an outcome widens residual variation, while error in a regressor can attenuate its slope under the lecture conditions.

In this chapter

What this chapter covers

  • 01

    Ordinary least squares

  • 02

    Fitted values and residuals

  • 03

    Multiple regression

  • 04

    R-squared

  • 05

    T-tests

  • 06

    Dummy interactions

Worked example · free

Recover an interaction slope

Q [5 marks]. The marks used here are not an official University assessment scheme. A model has a base slope of 0.60 and a positive-news interaction of 0.35. Find the positive-news slope and explain the interaction.
The fitted line shows the model prediction at each earnings-news value, while the vertical residual records the observation-specific prediction error.The fitted line shows the model prediction at each earnings-news value, while the vertical residual records the observation-specific prediction error.residualfitted lineearnings newsreturn
The fitted line shows the model prediction at each earnings-news value, while the vertical residual records the observation-specific prediction error.
  • 1Keep the base-group slope at 0.60 when the dummy equals zero.
  • 2Add 0.35 when the positive-news dummy equals one, giving 0.95.
  • 2Interpret 0.35 as the estimated difference between group slopes, not the entire positive-news response.
The positive-news slope is 0.95. The interaction of 0.35 measures how far that slope lies above the 0.60 base relation.
Sia tip — Substitute zero and one into a dummy model before using any coefficient label.
Glossary

Key terms

Ordinary least squares
An estimation method that selects coefficients to minimise squared residuals.
Regression residual
The observation's actual dependent value minus its fitted value.
Statistical precision
The degree of uncertainty around an estimated coefficient.
Interaction effect
A coefficient measuring how one relation differs across a defined group or condition.
FAQ

Regression Tools for Market Evidence FAQ

What should I read first in a table?

Identify the dependent variable, sample and base group before looking at significance symbols. Then decode the units of each regressor. This order prevents a statistically precise coefficient from being described as economically important without knowing what one unit means.

How is an interaction coefficient interpreted?

The interaction is normally a difference from the base relation. The full slope for the indicated group combines the base coefficient and interaction term. Its t-test may address the difference, while a separate calculation or joint test may be needed for the full group slope.

How does measurement error affect the model?

Noise in the dependent variable increases residual variation and weakens explanatory power. Error in an independent variable can bias its estimated slope toward zero under the lecture conditions. A good answer identifies which variable is noisy before predicting the consequence.

Can a small coefficient still matter?

Yes. Economic importance depends on the variable's scale and decision context, while statistical precision depends on uncertainty. Translate the estimate into stated units, report its test evidence separately, and avoid using significance as a substitute for magnitude.

Study strategy

Exam move

Annotate one fresh regression table each day. Circle the dependent variable, underline the base category, write every combined slope in the margin and translate one coefficient into the units printed in the question. Keep magnitude, precision and explanatory power in separate sentences. When evaluating a design, select one limitation that changes a specific variable or path instead of attaching a generic caveat.

Build fluency by inventing a small six-observation dataset and sketching a fitted line through it. Mark one fitted value and its vertical residual, then explain why OLS squares residuals before adding them. Change a single observation and predict whether fit should improve or worsen before calculating anything. On a separate sheet, make three headings: economic magnitude, statistical precision and explanatory power.

Force every table comment into the correct heading. A coefficient belongs under magnitude and direction; a t-statistic belongs under evidence about a null; R-squared belongs under the model's sample variation. Repeat this separation until significance language never replaces units. For dummy models, write the regression twice with the indicator set to zero and one.

Calculate both intercepts and slopes even when the question requests only one, because the unused calculation exposes sign errors. Use a miniature table to compare base, increment and combined effect. Add a sentence stating which test belongs to the difference and whether the full slope would need a separate test. Measurement-error revision should also be directional.

Draw noise entering the dependent variable and show larger residuals, then draw noise entering a regressor and explain attenuation under the maintained conditions. Avoid claiming every form of measurement error has the same consequence. Once a week, time a ten-minute interpretation of a new results table.

Spend the first two minutes mapping variables and base categories, the next five writing sign, size and evidence, and the final three adding one model-specific limitation. Review errors by type: unit omission, wrong base, uncombined interaction, causal overclaim or generic evaluation. That ledger reveals which reasoning step needs practice.

Working through Regression Tools for Market Evidence in BMAN30071? Sia is AskSia’s AI Accounting tutor — ask any BMAN30071 Regression Tools for Market Evidence question and get a clear, step-by-step explanation grounded in how BMAN30071 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.

A+Everything unlocked
Unlocks this Bible + all 41 of your subjects at The University of Manchester, plus 1,000+ more Bibles across the AskSia library.
Sia - your BMAN30071 tutor, unlimited, worked the way the exam marks it
The full 2-page Bible + practice bank with worked solutions
Chrome extension - sync your LMS so Sia knows your deadlines
Bilingual EN / Chinese on every Bible and every Sia answer
$0.99 Trial
30-day money-back · cancel in one tap · how it works