Ch. 8
Analytical Aptitude
The KSAOs needed to collect and analyze qualitative and quantitative data, and to interpret and promote findings that evaluate HR initiatives and inform business decisions and recommendations.
Sub-competencies
| Sub-competency | What it means |
|---|---|
| Data Advocate | Understanding and promoting the importance and usefulness of data |
| Data Gathering | Judging whether data is useful, and identifying and gathering data to inform decisions |
| Data Analysis | Analyzing data to evaluate HR initiatives and business challenges |
| Evidence-Based Decision-Making | Using the results of data analysis to decide the best course of action |
BASK key concepts: survey and assessment processes; data sources; data analysis techniques; basic statistics and measurement; interpreting graphs and charts; using data to support a business case.
Types of data and research
| Concept | Meaning |
|---|---|
| Quantitative data | Numerical data such as turnover rates, salaries and survey ratings |
| Qualitative data | Descriptive, non-numerical data such as interview notes, open-ended comments and focus group themes |
| Primary data | Data you collect yourself, such as your own survey |
| Secondary data | Data collected by others, such as BLS statistics and published salary surveys |
| Structured vs. unstructured data | Organized fields (HRIS records) vs. free text, video or e-mail |
| Big data | Very large, fast-moving, varied data sets that need advanced tools to analyze |
| Data cleaning / cleansing | Finding and fixing errors, duplicates, missing values and inconsistencies before analysis |
| Data mining | Using algorithms to discover patterns in large data sets |
| Predictive analytics | Using historical data and statistical models or machine learning to forecast outcomes, such as which employees are likely to quit |
| Benchmarking | Comparing your data or practices to other organizations or industry standards |
Data sources (BASK): interviews, focus groups, employee surveys, customer surveys, marketing data, analytical reports, payroll data, compensation and benefits data, employee and customer feedback, and employee KPI performance.
Descriptive statistics
| Measure | Meaning | Note |
|---|---|---|
| Mean | The arithmetic average (sum ÷ count) | Pulled toward extreme values (outliers) |
| Weighted mean | An average in which values count in proportion to their weight, such as the number of employees at each salary | Use it when groups differ in size |
| Median | The middle value when data is sorted. With an even count, it is the average of the two middle values. | Not distorted by outliers, so it is preferred for salary data |
| Mode | The most frequent value | Useful for categorical data |
| Range | Highest value − lowest value | Simple measure of spread |
| Standard deviation | Average distance of values from the mean | Larger means more spread out. In a normal (bell) curve, about 68% of values fall within ±1 SD of the mean and about 95% within ±2 SD. |
| Percentile | The value below which a given percentage of data falls | The 50th percentile is the median. Pay surveys report the 25th, 50th and 75th percentiles. |
| Quartiles | Values that divide sorted data into four equal parts | Q1 = 25th percentile, Q2 = median, Q3 = 75th percentile |
| Frequency distribution | A table or chart showing how often each value or range of values occurs | A histogram is its chart form |
Worked example
These are your course's salaries, sorted: $40,000; $45,000; $47,000; $47,000; $47,000; $47,500; $50,000; $58,000; $81,000; $81,000 (10 values).
- Mean = $543,500 ÷ 10 = $54,350
- Median = average of the 5th and 6th values = ($47,000 + $47,500) ÷ 2 = $47,250
- Mode = $47,000 (appears three times)
- Range = $81,000 − $40,000 = $41,000
The mean is higher than the median because the two $81,000 salaries pull it up. That is why the median better represents a "typical" salary.
Weighted mean example: 5 employees at $40,000, 2 at $45,000, 2 at $47,500, 1 at $50,000 and 5 at $81,000. The total is $840,000 for 15 employees, so the weighted mean is $56,000.
Relationships between variables
- Correlation measures the strength and direction of the relationship between two variables. The correlation coefficient (r) ranges from −1.0 (perfect negative) through 0 (no relationship) to +1.0 (perfect positive).
★ Correlation does not prove causation. A third factor may cause both.
- Regression analysis estimates how much one variable (the dependent variable) changes when one or more other variables (independent variables) change. It is used to predict outcomes, for example in pay equity analysis to control for legitimate pay factors.
- Scatter diagram (scatterplot): a visual of the relationship between two variables.
Analysis methods
| Method | Use |
|---|---|
| Variance analysis | Compares planned vs. actual results, such as a budget vs. actual spend |
| Ratio analysis | Compares financial statement figures to each other and to industry standards to assess financial health |
| Trend analysis | Examines data over time to identify patterns and forecast |
| Regression analysis | Measures the relationship between variables and predicts outcomes |
| Root cause analysis | Works backward from a problem to its fundamental cause. Techniques include the 5 Whys and the fishbone (Ishikawa / cause-and-effect) diagram. |
Charts and graphs (interpretation)
| Chart | Best for |
|---|---|
| Bar chart | Comparing categories, such as turnover by department |
| Line graph | Trends over time, such as monthly headcount |
| Pie chart | Parts of a whole (percentages adding to 100%) |
| Histogram | The distribution of a continuous variable in ranges, such as the number of employees in each salary band |
| Scatterplot | The relationship between two numerical variables |
| Pareto chart | Bars sorted from highest to lowest, plus a cumulative percentage line. Based on the 80/20 rule: about 80% of effects come from 20% of causes. It helps prioritize the "vital few." |
| Control chart | A process over time with upper and lower control limits, used to spot unusual variation (quality management) |
Measurement quality: reliability and validity
| Concept | Definition (BASK) | Types |
|---|---|---|
| Reliability | How far a measurement instrument provides consistent results | Test-retest (stability over time); inter-rater (agreement between raters); parallel or alternate forms; internal consistency (items measure the same thing) |
| Validity | How far an instrument measures what it is intended to measure | Content validity (it samples the actual job content); criterion-related validity (scores predict or correlate with performance; predictive or concurrent); construct validity (measures an abstract trait like leadership) |
★ A test can be reliable but not valid (it consistently measures the wrong thing). It cannot be valid without being reliable. Validity is what matters legally under the Uniform Guidelines on Employee Selection Procedures.
Survey and assessment processes
- Define objectives: what decision will this data inform?
- Design the survey: clear, unbiased, single-topic questions (avoid "double-barreled" questions), consistent rating scales (e.g., 5-point Likert scale) and a few open-ended questions.
- Pilot test it with a small group.
- Administer it with clear communication about confidentiality or anonymity.
- Analyze it, including open-ended response analysis (coding comments into themes).
- Report the results and act on them. Not acting reduces trust and future response rates.
Sampling: a random sample gives every member an equal chance of selection. A stratified sample ensures that subgroups are represented proportionally.
Measurement bias (rater errors)
| Bias | Description |
|---|---|
| Halo effect | One positive trait colors the whole evaluation |
| Horn effect | One negative trait colors the whole evaluation |
| Similar-to-me | Favoring people who share your characteristics or background |
| Stereotyping | Assuming that group generalizations apply to an individual |
| First impression / primacy | Early information outweighs later information |
| Recency | Recent events outweigh the full period |
| Contrast effect | Rating someone in comparison with other people rather than the standard |
| Central tendency | Rating everyone near the middle of the scale |
| Leniency / strictness | Rating everyone too high / too low |
| Negative emphasis | Over-weighting a small amount of negative information |
| Nonverbal bias | Over-weighting body language |
| Cultural noise | A candidate gives socially acceptable answers rather than true ones |
| Inconsistency | Applying different standards, questions or data collection to different people (e.g., selective interviewing) |
| Confirmation bias | Seeking out evidence that supports a pre-existing belief |
| AI / algorithmic bias | AI tools trained on biased historical data reproduce that bias. |
Evidence-based decision-making
Make decisions from the best available evidence, combining organizational data, scientific research, professional expertise and stakeholder perspectives. Tell the story with data: lead with the insight, use clear visuals, and connect it to the business impact.
Finished reading?