SPSS Biostatistics

Complete Answer Guide & Tutorial

Sample Exam · Total Marks: 40 · Duration: 2 Hours
Prepared for: Dr. Ashrafur Rahaman Mahadi | MPH (Epidemiology), CUB

Question 1 – Conceptual Knowledge (10 Marks)

1(a) Key Definitions with Examples

Any measurable characteristic that differs between individuals or observations. →

Variable
Example: GPA, Age, Gender — all differ from student to student.

Independen The predictor/exposure variable. The one we hypothesize causes or explains changes
t Variable in the outcome. → Example: Study Hours — we expect it to predict GPA.
(IV)

Dependent The outcome/response variable. The one we measure as the effect. → Example: GPA
Variable — it depends on study hours.
(DV)

Scale of Measurement
Scale Definition Example from Dataset

Nominal Categories with NO order; just Gender (Male/Female), Smoking (Yes/No)

labels

Ordinal Categories WITH order; intervals Satisfaction (Low < Medium < High)
NOT equal

Scale (Ratio) Numeric, equal intervals, true Age, GPA, Study Hours, Attendance%
zero exists

1(b) When to Use Which Regression?

Regression Type Use When DV is… Example

Linear Regression Continuous (Scale/Ratio) Predicting GPA from Study Hours

Logistic Regression Binary (Yes/No, 0/1) Predicting Smoking status (Yes/No)

 Regression Type Use When DV is… Example

Ordinal Regression Ordered categories (Ordinal) Predicting Satisfaction (Low/Med/High)

1(c) What is R²?

R² (R-squared) = Coefficient of Determination. It tells us what proportion (%) of the
Definition variance in the DV is explained by the IV(s). Formula: R² = (Explained Variation) ÷
(Total Variation)

Interpretation: 60% of the variation in GPA can be explained by the independent

R² = 0.60 variables in the model. The remaining 40% is due to other factors not included in the
model. → This is considered a GOOD model fit in social/health sciences.

1(d) What is Odds Ratio (OR)?

OR = the ratio of the odds of an event occurring in one group vs. another group. OR =
Definition
(odds of outcome in exposed) ÷ (odds of outcome in unexposed)

OR Value Interpretation

OR = 1 No difference between groups (null effect)

OR > 1 Higher odds of outcome in exposed group → RISK FACTOR

OR < 1 Lower odds of outcome in exposed group → PROTECTIVE FACTOR

Question 2 – Data Setup & Univariate Analysis (10 Marks)

Dataset Used Throughout This Exam

ID Age Gender Study_Hrs Attendance GPA Smoking Satisfactio
% n
1 20 Male 2 60 2.5 Yes Low
2 21 Female 3 65 2.7 No Low
3 22 Male 4 70 3.0 Yes Medium
4 23 Female 5 75 3.2 No Medium
5 24 Male 6 80 3.5 Yes High
6 25 Female 7 85 3.7 No High
7 26 Male 8 90 3.8 Yes High
8 27 Female 9 95 4.0 No High
2(a) Variable Type & Scale of Measurement
Variable Type Scale of Measurement SPSS Example Value
Measure

ID Discrete Numeric Nominal (identifier) Nominal 1, 2, 3…

Age Continuous Numeric Scale (Ratio) Scale 20, 21, 22

Gender Categorical Nominal Nominal Male / Female

Study_Hours Continuous Numeric Scale (Ratio) Scale 2, 3, 4

Attendance (%) Continuous Numeric Scale (Ratio) Scale 60, 70, 80

GPA Continuous Numeric Scale (Ratio) Scale 2.5, 3.0

Smoking Categorical (Binary) Nominal Nominal Yes / No

Satisfaction Categorical (Ordered) Ordinal Ordinal Low/Medium/

High

2(b) SPSS Data Entry – Step by Step

Variable View Setup
1 Open SPSS → click 'Variable View' tab (bottom left of screen)

2 For each variable, fill in: Name (no spaces), Type (Numeric / String), Width, Decimals, Label,
Values, Measure

3 ID → Numeric, Decimals: 0, Measure: Nominal

4 Age → Numeric, Decimals: 0, Measure: Scale

5 Gender → String (or Numeric with Value Labels: 1=Male, 2=Female), Measure: Nominal

6 Study_Hours → Numeric, Decimals: 0, Measure: Scale

7 Attendance → Numeric, Decimals: 1, Measure: Scale

8 GPA → Numeric, Decimals: 2, Measure: Scale

9 Smoking → Numeric, Value Labels: 1=Yes, 0=No, Measure: Nominal

10 Satisfaction → Numeric, Value Labels: 1=Low, 2=Medium, 3=High, Measure: Ordinal

Data View Entry

1 Click 'Data View' tab

2 Enter data row by row — each row = one participant (N=8)

3 For categorical variables, enter the numeric code (e.g., 1 for Male, 0 for No)

4 Verify: 8 rows × 8 columns after entry

5 Save as: File → Save As → [Link]

2(c) Univariate Analysis in SPSS
For Continuous Variables (Age, Study Hours, Attendance, GPA)
1 Go to: Analyze → Descriptive Statistics → Descriptives

2 Move continuous variables to 'Variables' box

3 Click Options → check: Mean, Std. Deviation, Minimum, Maximum → Continue → OK

For Categorical Variables (Gender, Smoking, Satisfaction)

1 Go to: Analyze → Descriptive Statistics → Frequencies

2 Move categorical variables to 'Variables' box

3 Click Statistics → check: Mode → Continue

4 Click Charts → Bar charts → OK

Expected Output Summary (Calculated from Dataset)

Variable Mean Std. Deviation N

Age 23.5 2.45 8

Study Hours 5.5 2.45 8

Attendance (%) 77.5 12.22 8

GPA 3.24 0.54 8

Categorical Variable Category Frequency (%)

Gender Male 4 (50%)

Female 4 (50%)

Smoking Yes 4 (50%)

No 4 (50%)

Satisfaction Low 2 (25%)

Medium 2 (25%)

High 4 (50%)

2(d) Interpretation
The average GPA of the 8 students is 3.24 out of 4.0, indicating above-average
Mean GPA
academic performance overall. The SD of 0.54 shows moderate variability among
= 3.24
students.
 50% of students reported High satisfaction, 25% Medium, and 25% Low. The
Satisfaction
distribution is positively skewed — more students are satisfied than unsatisfied. This
Distribution
aligns with the higher GPA pattern in the dataset.

Question 3 – Bivariate Analysis & Simple Linear Regression (10

Marks)

3(a) Hypotheses

H₀ Null Hypothesis: There is NO significant relationship between Study Hours and GPA. (ρ =
0)

H₁ Alternative Hypothesis: There IS a significant relationship between Study Hours and

GPA. (ρ ≠ 0)

3(b) Pearson Correlation in SPSS

1 Analyze → Correlate → Bivariate

2 Move Study_Hours and GPA to 'Variables' box

3 Ensure 'Pearson' is checked under Correlation Coefficients

4 Check 'Flag significant correlations' → OK

Expected Pearson r ≈ 0.997 (very strong positive correlation) p-value < 0.001 (highly significant)
Result Interpretation: As Study Hours increase, GPA increases almost perfectly. Reject H₀.

3(c) Simple Linear Regression in SPSS

1 Analyze → Regression → Linear

2 Move GPA to 'Dependent' box

3 Move Study_Hours to 'Independent(s)' box

4 Click Statistics → check: Estimates, Confidence intervals, Model fit → Continue

5 Click OK to run

3(d) Interpreting Regression Output

Output Statistic Value (Expected) What It Means Decision

R² 0.994 99.4% of GPA variance Excellent model fit

explained by Study Hours
 Output Statistic Value (Expected) What It Means Decision

Beta (B) coefficient ≈ +0.214 For each 1 extra study Positive relationship
hour, GPA increases by
0.214

Constant (a) ≈ 2.07 GPA when Study Hours = Baseline

0

p-value (Study Hours) < 0.001 Study Hours is a Reject H₀

SIGNIFICANT predictor
of GPA

Regression GPA = 2.07 + (0.214 × Study Hours) Example: A student studying 6 hours/day → GPA
Equation = 2.07 + (0.214 × 6) = 2.07 + 1.284 = 3.35

Question 4 – Multivariate Analysis (10 Marks)

Part A: Multiple Linear Regression

4A(a) Regression Model Equation

GPA = β₀ + β₁(Study Hours) + β₂(Attendance%) + β₃(Age) + ε Where: β₀ =
MLR
constant (intercept) β₁, β₂, β₃ = regression coefficients for each predictor ε = error
Formula
term (residual)

4A(b) Running MLR in SPSS & Interpretation

1 Analyze → Regression → Linear

2 Dependent: GPA

3 Independents: Study_Hours, Attendance, Age

4 Click Statistics → check: Estimates, Confidence intervals, Model fit, Collinearity diagnostics →
Continue → OK

Predictor Expected Beta p-value Interpretation

Study Hours +0.15 < 0.05 (Sig.) Significant positive predictor; more
study → higher GPA

Attendance (%) +0.01 < 0.05 (Sig.) Significant; higher attendance →

higher GPA

Age +0.05 > 0.05 (NS) Not significant in this model; age
doesn't independently predict GPA

Adjusted R² Adjusted R² accounts for the number of predictors in the model. Expected value ≈ 0.99
 → The model explains ~99% of GPA variance. Adjusted R² is preferred over R² in MLR
to avoid overestimation.

Part B: Ordinal Regression

4B(a) Why NOT Linear Regression for Satisfaction?

Linear regression assumes: 1. The DV is continuous (Satisfaction is categorical) 2.
Equal intervals between categories (Low→Medium ≠ Medium→High in reality) 3.
Problem
Residuals are normally distributed Satisfaction (Low/Medium/High) violates all these
with Linear
assumptions. Using linear regression would treat 'Low=1, Medium=2, High=3' as if the
Regression
difference between each level is equal and meaningful on a ratio scale — which is
WRONG.

4B(b) Why Ordinal Regression is Correct

Ordinal regression (Proportional Odds Model): ✓ Respects the ORDER of categories
Advantages (Low < Medium < High) ✓ Does NOT assume equal intervals between categories ✓
of Ordinal Produces odds ratios for each predictor ✓ Uses cumulative logits: compares P(Y ≤ j)
Regression vs P(Y > j) at each threshold This is the statistically correct model when the DV is
ordinal.

4B(c) The Ordinal Regression Model – Concept

logit[P(Y ≤ j)] = αⱼ – β₁(Study Hours) – β₂(Attendance) Where: • αⱼ = threshold
parameters (one for each cut-point: Low|Medium and Medium|High) • β = regression
Model
coefficients (negative sign because SPSS uses the 'reverse' coding by default) • The
Equation
model estimates the cumulative odds of being in category ≤ j Key Assumption:
Proportional Odds — the effect of each predictor is CONSTANT across all thresholds.

1 Analyze → Regression → Ordinal

2 Dependent: Satisfaction

3 Covariates: Study_Hours, Attendance (Scale variables go here)

4 Link Function: Logit (default — appropriate for ordinal outcomes)

5 Click Output → check: Parameter estimates → Continue → OK

4B(d) Interpreting Ordinal Regression Output

Parameter Estimate (β) Exp(β) = OR Interpretation

[Satisfaction = 1] Low| α₁ (threshold) — Cut-point between Low and Medium

Med satisfaction

[Satisfaction = 2] Med| α₂ (threshold) — Cut-point between Medium and High

High satisfaction

Study Hours -0.8 (example) OR ≈ 0.45 Each 1 hr increase in study → 55% lower
odds of being in lower satisfaction
Parameter Estimate (β) Exp(β) = OR Interpretation

category → Higher satisfaction

Attendance (%) -0.05 (example) OR ≈ 0.95 Each 1% increase in attendance →

slightly lower odds of lower satisfaction

Key SPSS ordinal regression uses NEGATIVE coefficients by default (it models the odds of
Interpretati being in LOWER categories). So: A NEGATIVE β means HIGHER predictor value →
on Rule for HIGHER satisfaction (better outcome). OR < 1 for a predictor: as that predictor
Ordinal increases, the odds of being in a lower satisfaction group decrease → i.e., satisfaction
Regression improves.
in SPSS

Quick Reference Card – SPSS Menu Paths

Analysis SPSS Menu Path

Frequencies (Categorical) Analyze → Descriptive Statistics → Frequencies

Descriptives (Continuous) Analyze → Descriptive Statistics → Descriptives

Pearson Correlation Analyze → Correlate → Bivariate → ✓ Pearson

Simple Linear Regression Analyze → Regression → Linear (1 IV, 1 DV)

Multiple Linear Regression Analyze → Regression → Linear (multiple IVs, 1 DV)

Ordinal Regression Analyze → Regression → Ordinal → Link: Logit

Logistic Regression Analyze → Regression → Binary Logistic

Chi-Square Test Analyze → Descriptive Statistics → Crosstabs → ✓ Chi-square

Independent t-test Analyze → Compare Means → Independent Samples T-test

ANOVA Analyze → Compare Means → One-Way ANOVA

End of Answer Guide · Total Marks: 40 · Good Luck!