π Lecture 24 β Fieldwork
π Overview: This lecture covers the critical process of fieldwork in marketing research, emphasizing that poor management of this stage can neutralize all previous research efforts. It details the four main aspects of fieldworkβtime schedule, budget, personnel, and performance measurementβand provides comprehensive guidance on selecting, training, and managing the field force to ensure high-quality data collection.
ποΈ Topics Covered
The lecture begins by defining fieldwork and its importance in data collection, then outlines the four main aspects: time schedule, budget, personnel, and performance measurement. It delves into details of creating realistic time schedules and budgets, then extensively covers personnel management including selection criteria, training requirements, probing techniques, and proper procedures for recording replies and terminating interviews.
π Lecture Summary
Fieldwork
Fieldwork is an important step in the research process equated to data collection and related activities such as contacting respondents, administering instruments, recording data, and returning data to a central location for processing. If it is poorly managed, all previous steps will be neutralized. Collection of data may be by telephonic, face-to-face, or mail methods, and the field design must be organized accordingly, including activities of supervision, monitoring, mailing, and observing all field operations. If fieldwork is not properly organized, it will result into many kinds of errors. The four main aspects of fieldwork are: time schedule, budget, personnel, and performance measurement.
π‘ Why this matters: Proper fieldwork management is the foundation upon which the validity of all research findings rests; errors at this stage cannot be corrected later.
Time Schedule
When preparing a time schedule of activities, we must be realistic and allocate a reasonable amount of time to data collection within the total time budget. To make a time schedule, the planner should specify the beginning and ending of the project, sequence activities within the time frame, determine days needed to complete various activities, and prepare a Gantt chart.
Budget
Budget includes assignment of cost to various activities identified in the time schedule. Both the time schedule and budget are prepared together. The budget shows a detailed breakdown of the cost of all activities and includes some amount for unseen contingencies. It is reviewed and approved by some individuals in the organization. Various cost items include: wages and salaries, telephone expenses, material and supplies, salaries of field supervisor/s, reproduction expenses like photocopy, and miscellaneous.
Personnel
Personnel means human resources for the fieldwork, collectively called the field force. Management of personnel requires selection, training, supervising, authenticating, and evaluation of field workers. For selection, we have to develop Job Description and Job Specification. The characteristics of the field force should include: being healthy and having stamina to undertake strenuous work, having a pleasant appearance, being communicative with listening and speaking skills, being educated up to BA, and preferably having some experience. Inexperienced personnel result in excessive coding error, misreporting, and larger refusal rate.
Training of Field Force
After personnel have been recruited, they should be trained in the skills of data collection. Training is critical for fieldworkers and should be given in person at some central location. It should cover: initial contact, appointments, and opening remarks to seek permission and cooperation; training about observation; asking questions, which brings high dividends in eliminating potential bias; being thoroughly familiar with the questionnaire; observing the order of the questions; using exact wording; slow reading to help respondents understand the question; and repeating questions if not understood. Probing should be done carefully.
Probing Questions
Probing questions are used to motivate, clarify, enlarge, or explain the answers. Techniques include: repeating the question in the same words; repeating the reply verbatim and confirming; silent probe by unexpected pause and looks that should not be embarrassing; reassuring and boosting the respondent that there is no right or wrong, just what it means to him/her; and eliciting clarification by saying "I don't quite understand", "please tell me more", "what do you mean", "anything else", "any other reason?"
Recording Replies
Recording replies seems simple but mistakes are common. Unstructured questions must be recorded verbatim. Guidelines for recording include: record identification data like name, date, interviewer, and project in the beginning; record during the interview and not afterward; record in respondents' own words; do not summarize; include everything; include all probe comments; and repeat the response so that respondent listens and verifies.
Terminating Interview
Guidelines for closing the interview include: don't close before all information is gathered; leave the respondent with good feelings; and express your appreciation for his time and cooperation. In telephonic interviews: verify the telephone number once again; be courteous and polite; put a smile in your voice; speak clearly; don't sound bored; and close the interview properly.
β Key Takeaways
The four critical aspects of fieldwork management are time schedule, budget, personnel, and performance measurement, and these must be carefully coordinated to avoid errors that can invalidate the entire research project. Proper selection of field force members requires specific characteristics including health, appearance, communication skills, education, and experience, as inexperienced personnel lead to excessive errors and refusal rates. Training must cover initial contact procedures, question administration with exact wording and proper order, and careful probing techniques. Interviewers must record replies verbatim during the interview using respondents' own words without summarizing, and must terminate interviews professionally by ensuring all information is gathered and leaving respondents with positive feelings.
π§ Quick Revision Questions
- What are the four main aspects of fieldwork that must be managed to avoid research errors?
- What specific characteristics should a field force member possess according to the job specification?
- List at least five techniques that can be used for probing questions during interviews.
- What are the key guidelines for properly recording replies during data collection?
- How should an interviewer properly terminate an interview, and what special considerations apply to telephonic interviews?
π Lecture 25 β Data Analysis
π Overview: This lecture covers the critical processes of performance measurement of field force, data preparation, and data editing in marketing research. It emphasizes the importance of controlling non-sampling errors that occur during fieldwork, distinguishing between errors committed by fieldworkers and respondents, and providing practical methods to minimize these errors.
ποΈ Topics Covered
The lecture begins with performance measurement of the field force, including authentication and evaluation methods. It then transitions into data preparation and the principles of data analysis, followed by data editing procedures to fix common problems. A major portion is dedicated to non-sampling errors, categorized into fieldworker errors (intentional and unintentional) and respondent errors (intentional and unintentional), concluding with detailed strategies for controlling each type of error.
π Lecture Summary
Performance Measurement of the Field Force
Performance measurement and management are two important aspects of fieldwork. The supervisor should ensure that procedures are being followed; if not, provide additional training to the field workers. Guidelines include examining and editing collected questionnaires, ensuring the sampling plan is followed strictly (not based on convenience and accessibility), and controlling cheating and fake answers by calling respondents on the phone.
Authentication & Evaluation A good method to authenticate that the data were collected genuinely is to call 10 to 15% of respondents to validate. Ask their basic demographic data and cross-check. Also ask respondents about the length, quality, and their reaction to the interviewer. To evaluate the performance of each field worker, check: contact rate, response rates, refusal rate, cost incurred and time spent, quality of interviewing (including precision, ability to probe, ability to ask sensitive questions, interpersonal skills, and termination of interview), and quality of data (legibility, non-response, whether instructions were followed, and if answers were completely recorded).
Data Preparation
We should understand that data is recorded measures of phenomenon, but information is a body of facts in a format suitable for decision making. The purpose of research is to provide information. Data analysis is the process whereby data is converted into information. Understanding the principles of data analysis is important because it: leads the researcher to develop insight into information and data, helps avoid erroneous judgment and conclusions, and helps interpret the analysis of others.
π‘ Why this matters: Knowledge and power of data analysis can constructively influence research design, but it cannot rescue or compensate for a study that was not well conceived. If research hypotheses were non-viable, sampling was inadequate, or fieldwork was sloppy, data analysis will not provide any remedy.
Data Editing
Although data analysis techniques are unique for each study, all studies require data editing. Data editing means identifying omissions, ambiguities, and errors in the responses and taking necessary corrective actions. The responsibility for these errors could lie with the interviewer, supervisor, or data analyst.
Problems identified during editing include: incorrect instructions by the interviewer; omissions and ambiguities (e.g., two boxes checked in an MCQ); inconsistencies (e.g., not married but having children); lack of cooperation (always agreeing); and ineligible respondents (e.g., age under 18). Alternatives for data editing include: approaching the respondent again for clarification, devising a new category, creating a category of "no answer" or "missing" , or not including a question or the whole questionnaire in the analysis.
Non Sampling Errors
All errors in a survey, except those related to the sampling plan and sampling size, are non-sampling errors. These include: non-response errors, data gathering errors, data handling errors, data analysis errors, interpretation errors, ambiguity in problem definition, and inappropriate wording of questions. The greatest potential for these errors is at the data collection stage or fieldwork. Such errors may occur due to the fieldworker or the respondents.
Fieldworker Errors These errors are committed by the person who administers the questionnaire or takes the interview. They may be intentional (committed deliberately) or unintentional (occur without willful intent).
- Intentional Fieldworker Errors: Cheating (e.g., intentionally falsifying responses if paid per completed interview), contacting a convenient interviewee instead of a genuine sample, completing fewer questionnaires than agreed, leading the respondent to a particular answer through wording or body language, and re-wording the question.
- Unintentional Fieldworker Errors: These arise from three sources: personal characteristics (accent, gender), misunderstanding of instructions about scales or recording responses, a gap between the educational levels of the research designer and the fieldworker, and miscellaneous sources like fatigue or monotony.
Respondent Errors These errors are committed by the person who fills the questionnaire or responds to an interview. They are also either intentional or unintentional.
- Respondents' Intentional Errors: Respondents willfully misrepresent themselves by telling a lie (due to privacy or embarrassment regarding income, age, etc.) or by not giving a response in whole or part (due to a busy schedule or privacy).
- Respondents' Unintentional Errors: Respondents believe an invalid response is true. Examples include answering without understanding (e.g., income with or without taxes), checking two answers instead of one, guessing due to low recall, attention loss, distraction, and respondent fatigue.
How to Control Field Errors
Precautions should be taken to minimize such errors, but they cannot be eliminated entirely. Several methods are suggested for controlling field errors.
Control of Intentional Fieldworker Errors:
- Supervision: Oversee the work, use central telephone monitoring to spot cheating, and reprimand the interviewer if needed.
- Validation: Validate the work by re-contacting about 10% of respondents, sometimes re-administering the instrument for comparison.
Control of Unintentional Fieldworker Errors:
- Selection and Training: Careful selection and training can best care for such errors.
- Orientation Session: Meetings of field workers with the supervisor to give instructions.
- Role Playing (Rehearsals): The supervisor or another person plays the role of respondent to check interviewing skills.
Control of Intentional Respondent Errors:
- Anonymity: Assure confidentiality of the respondent's name.
- Incentives: Give or promise gifts like t-shirts, ballpoints, cash, or diaries.
- Validation Checks: Confirm the response by cross-checking with other ways.
- Third Person Techniques: Design questions so they are indirect, in the name of a friend or neighbor.
Control of Unintentional Respondent Errors:
- Well Drafted Questionnaire: Clear instructions and examples remove misunderstandings.
- Reversal of Scale on Endpoints: Place negative adjectives on different sides (sometimes right, sometimes left) to compel thoughtful answers.
- Prompters: Use written or oral statements like "we are almost finished" to keep respondents on task and reduce fatigue.
π‘ Why this matters: Good design of the questionnaire and technology can reduce errors of data collection in general.
β Key Takeaways
Data analysis converts raw data into decision-making information, but it cannot fix a poorly designed study. All field measurement errors, from both fieldworkers and respondents, are non-sampling errors which must be actively controlled. Intentional errors like cheating can be managed through supervision and validation, while unintentional errors require better training, clear instructions, and well-drafted questionnaires. Authenticating 10-15% of respondents is a critical step to ensure data validity and to measure field force performance.
π§ Quick Revision Questions
- What is the difference between "data" and "information" as defined in the lecture?
- What are the five alternatives for data editing when a problem is found in a questionnaire?
- List the three methods for controlling unintentional fieldworker errors.
- What is the "third person technique" and which type of error is it designed to control?
- Why is it important to reverse scale endpoints on a questionnaire?
π Lecture 26 β Data Analysis
π Overview: This lecture introduces the process of data analysis in marketing research, emphasizing that it is as important as other research phases. It covers the distinction between descriptive and inferential statistics, and explains basic data analysis techniques including frequency distribution, measures of central tendency (mean, median, mode), and their appropriate applications based on data type and measurement levels.
ποΈ Topics Covered
The lecture begins by positioning data analysis within the broader marketing research process, then categorizes statistical techniques into descriptive and inferential statistics. It explains how variable count and measurement levels influence technique selection. The main focus is on basic data analysis procedures: frequency distribution with examples, arithmetic mean calculation for interval data (both raw and grouped), median as a measure for ordinal data with odd/even observation calculations, and mode as the most frequent observation.
π Lecture Summary
Data Analysis Overview
Data analysis proceeds after data have been collected and checked for legibility, completeness, consistency, accuracy, and response classification. It is one of many activities in marketing research but not the most important aspect; rather, it is as important as other aspects such as good problem definition, appropriate study design, proper sampling procedures, well-designed and tested instruments of data collection, and well-monitored fieldwork. In data analysis, the researcher needs adequate knowledge of analysis techniques and must select techniques aligned with research objectives.
Categories of Statistical Techniques
Statistical techniques fall into two categories: descriptive statistics and inferential statistics. Descriptive statistics describe the characteristics of the sample under study, while inferential statistics help the researcher make inferences about the population from which the sample was drawn. Selection of statistical techniques also depends on how many variables are to be analyzed at the same time: uni-variable, bi-variable, and multi-variable analysis. Another relevant question is what level of measurement is available: nominal, ordinal, or interval.
Basic Data Analysis Procedures
Three basic data analysis procedures discussed are frequency distribution, cross-tabulation, and hypothesis testing.
Frequency Distribution
A frequency distribution simply reports the number of responses that each question received. When a market researcher is finding an answer to questions about a single variable, they start with frequency distribution. Absolute frequency and relative frequency (percentage) may be the subject of frequency distribution.
π Definition β Frequency Distribution: Reports the number of responses each question received.
π Example 1 β Distribution of customers of brand X of the 1300cc car province wise (n=780):
- Islamabad: 034 (04.4%)
- Punjab: 305 (39.1%)
- Sindh: 230 (29.5%)
- NWFP: 148 (19.0%)
- Baluchistan: 063 (08.0%)
- Total: 780 (100%)
π Example 2 β Factors important in purchase of Refrigerator (n=300):
- Size: 183 (61%)
- Color, appearance: 90 (30%)
- Price: 57 (19%)
- Less consumption of electricity: 90 (30%)
- Size of freezer: 54 (18%)
- Self defrosting: 60 (20%)
- Ice maker, Water dispenser: 84 (28%)
- Brand name: 51 (17%)
π Example 3 β Monthly income of respondents (n=200):
- 0-5000: 15 (7.5%)
- 5000-8000: 22 (11.0%)
- 8000-11000: 42 (21.0%)
- 11000-14000: 60 (30.0%)
- 14000-17000: 33 (16.5%)
- 17000-20000: 28 (14.0%)
- Total: 200 (100%)
Mean
Mean is a measure of central tendency. There are different kinds of means, e.g., arithmetic mean, geometric mean, harmonic mean. In marketing research, the arithmetic mean (arithmetic average) is most widely used. It is the ratio between the sum of observations and number of observations. Mean is the sum of values divided by the sample size or number of observations.
π Formula: Mean = Sum of values / Sample size (number of observations)
π Example β Raw data (n=10): Family # Monthly Income ('000 rupees')
- 1: 41
- 2: 37
- 3: 28
- 4: 51
- 5: 50
- 6: 36
- 7: 30
- 8: 41
- 9: 47
- 10: 29 Sum: 390 Mean Income = 390/10 = 39 (or rupees 39,000)
If the data are interval grouped into categories and classes, mean is calculated by taking the midpoints of the categories. The midpoint of a category is multiplied by the frequency in the category, summing up the multiplied values and dividing by the sum of frequencies.
π Example β Grouped data (n=200): Monthly Income (Rs.) | # of Respondents (f) | Midpoint (x) | fx 0-5000 | 15 | 2500 | 37,500 5000-8000 | 22 | 6500 | 143,000 8000-11000 | 42 | 9500 | 399,000 11000-14000 | 60 | 12,500 | 750,000 14000-17000 | 33 | 15,500 | 511,500 17000-20000 | 28 | 18,500 | 518,000 Total | 200 | | 2,359,000 Mean income = 2,359,000 / 200 = Rs. 11,795
Median (for Ordinal Data)
In ordinal data, median is the measure of central tendency. Median is the midpoint of the frequencies. As per this measure, 50% of the observations are above median value whereas 50% are below.
π Definition β Median: The midpoint of frequencies such that 50% of observations are above and 50% are below.
If the number of observations is odd, the middle observation is the median.
π Example β Odd number of observations: Sales values by a salesman in a month: Rs. 125,000, Rs. 175,000, Rs. 325,000, Rs. 77,000, Rs. 180,000. Arrange in ascending order: Rs. 77,000 Rs. 125,000 Rs. 175,000 Rs. 180,000 Rs. 325,000 The median (middle value) is Rs. 175,000.
If the observations are even numbers, the median would be the arithmetic mean of the middlemost two values.
π Example β Even number of observations: Sales order values in a month: Rs. 25,000, Rs. 178,000, Rs. 510,000, Rs. 450,000, Rs. 275,000, Rs. 370,000. Arrange in descending order: Rs. 510,000 Rs. 450,000 Rs. 370,000 Rs. 275,000 Rs. 250,000 Rs. 178,000 Middlemost two values: Rs. 370,000 and Rs. 275,000 Median = (370,000 + 275,000) / 2 = 645,000 / 2 = Rs. 322,500
Mode
Mode is still another measure of central tendency. This is the observation (x) where the frequency is maximum.
π Definition β Mode: The observation where the frequency is maximum.
π Example β Sales returns by model: Model # | 1 | 2 | 3 | 4 | 5 | 6 | 7
of returns | 6 | 12 | 8 | 2 | 1 | 7 | 2
The mode of sales return is Model #2 because the number of returns was 12 (maximum frequency).
β Key Takeaways
Data analysis is as important as other marketing research phases and requires selecting appropriate techniques based on research objectives, number of variables, and measurement levels. The key distinction is between descriptive statistics (describing sample characteristics) and inferential statistics (making population inferences). For single-variable analysis, frequency distribution reports response counts and percentages. The arithmetic mean is the most widely used measure of central tendency for interval data, calculated as sum of values divided by number of observations; for grouped data, midpoints are used. Median is appropriate for ordinal data and is the midpoint of frequencies, calculated differently for odd (middle value) versus even (average of two middle values) numbers of observations. Mode identifies the most frequent observation in a dataset.
π§ Quick Revision Questions
- What are the two main categories of statistical techniques in marketing research, and how do they differ?
- How do you calculate the mean for grouped interval data, and what is the formula?
- What is the difference between calculating median for an odd versus even number of observations?
- In the example of factors important in refrigerator purchase, what was the most frequently cited attribute and its percentage?
- What measure of central tendency is most appropriate for ordinal data, and why?
π Lecture 27 β Measure of Dispersion
π Overview: This lecture moves beyond measures of central tendency to explore measures of dispersion, which describe the variation or spread in data. Understanding dispersion is crucial for market researchers because two datasets can have identical means but very different distributions, affecting business decisions like distribution planning. The lecture covers range, variance, standard deviation, and coefficient of variation, including how to compute and interpret them.
ποΈ Topics Covered
The lecture begins by explaining why measures of dispersion are needed, using an example of two household groups with identical mean tea spending but different spreads. It defines Range as the simplest measure, then explains Variance as the mean of squared deviations from the mean, and Standard Deviation as its square root. A detailed example using ABC Real Estate shows how to compute sample variance and standard deviation step-by-step. The Rules of Standard Deviation for normal distributions are presented, followed by a bank waiting-time example that calculates all measures and evaluates a manager's statement. Finally, the Coefficient of Variation (CV) is introduced as a relative measure of dispersion for comparing datasets with different magnitudes, illustrated with a stock investment example.
π Lecture Summary
Why Measure Dispersion?
Measures of location (mean, mode, median) describe central tendency, but market research also needs to know the variation or scatter among values. Two variables can have identical central tendencies but very different spread. For example, spending on tea per annum for two groups of households: Group A (3000, 1500, 2100, 1800, 2800) and Group B (3600, 972, 1319, 2700, 2609) both have a mean of Rs. 2240, but their variation is not the same. This information is important for a tea manufacturer making a distribution plan.
Range
The range is the difference between the largest and smallest values in the data. In the previous example, the range of Group A is Rs. 3000 β 1500 = 1500, while the range of Group B is Rs. 3600 β 972 = 2628.
Variance and Standard Deviation
The distance between the mean and an observed value is the deviation from the mean. When we square these deviations and find the mean of these squared deviations, this is called variance. When data are scattered to a great extent, variance is large; if data are clustered around the mean, variance is small.
π Definition β Sample Variance (sΒ²): It is found with the formula ( s^2 = \frac{\sum (x - \bar{x})^2}{n - 1} ), where the sum of square deviations for the value from the mean ( \bar{x} ) is divided by ( n - 1 ).
π Definition β Standard Deviation of Sample (s): It is the square root of the sample variance. The formula is ( s = \sqrt{\frac{\sum (x - \bar{x})^2}{n - 1}} ).
π Example β ABC Real Estate in Karachi wants to know how long it takes to sell listed homes. The director took a sample of 10 homes and recorded weeks to sell: 21, 6, 9, 23, 1, 10, 8, 11, 5, 7.
| Home | Period x | (x - (\bar{x})) | (x - (\bar{x}))Β² |
|---|---|---|---|
| 1 | 21 | 10.9 | 118.81 |
| 2 | 6 | -4.1 | 16.81 |
| 3 | 9 | -1.1 | 1.21 |
| 4 | 23 | 12.9 | 166.41 |
| 5 | 1 | -9.1 | 82.81 |
| 6 | 10 | -0.1 | 0.01 |
| 7 | 8 | -2.1 | 4.41 |
| 8 | 11 | 0.9 | 0.81 |
| 9 | 5 | -5.1 | 26.01 |
| 10 | 7 | -3.1 | 9.61 |
| n=10 | (\sum x=101) (\bar{x}=101/10=10.1) | (\sum = 426.9) |
( s = \sqrt{\frac{426.9}{10 - 1}} = \sqrt{\frac{426.9}{9}} = \sqrt{47.433} \approx 6.89 ) weeks.
Rules of Standard Deviation
For a normal distribution: 68% of the population falls within Β±1 standard deviation. 95% of the population falls between Β±2 standard deviations. 99.7% (virtually all) of the population falls between Β±3 standard deviations.
π Example β A bank branch records waiting time (in minutes) during lunch hour for 12 customers: 4, 5, 3, 5, 6, 2, 7, 2, 4, 3, 2.
Solution:
| Customer # | Waiting time x | (x - (\bar{x})) | (x - (\bar{x}))Β² |
|---|---|---|---|
| 1 | 4 | 0 | 0 |
| 2 | 5 | 1 | 1 |
| 3 | 3 | -1 | 1 |
| 4 | 5 | 1 | 1 |
| 5 | 5 | 1 | 1 |
| 6 | 6 | 2 | 4 |
| 7 | 2 | -2 | 4 |
| 8 | 7 | 3 | 9 |
| 9 | 2 | -2 | 4 |
| 10 | 4 | 0 | 0 |
| 11 | 3 | -1 | 1 |
| 12 | 2 | -2 | 4 |
| n=12 | (\sum x=48) (\bar{x}=48/12=4) | (\sum = 1.82) |
Mean = 4 minutes; Median: arranging in descending order (7,6,5,5,5,4,4,3,3,2,2,2), median = (4+4)/2 = 4; Mode = 5; Variance (sΒ²) = 3.312; s = 1.82. The manager's statement "almost certainly not more than six minutes" is correct because the range of Β± one standard deviation is (4-1.82) to (4+1.82) = 2.18 to 5.82 minutes. Thus, majority of people wait not more than 5.82 minutes, which supports the manager's claim.
π‘ Why this matters: The standard deviation rule allows managers to make probabilistic statements about expected customer experience, such as waiting times, which is valuable for service quality management.
Coefficient of Variation
The coefficient of variation (CV) is the ratio of the standard deviation to the mean, expressed as a percentage: ( CV = \frac{s}{\bar{x}} \times 100 ). CV is expressed in percentage and is useful when the variable is measured on a ratio scale. It is a useful measure of relative dispersion when means are positive, allowing comparison of sets of numbers with different magnitudes.
π Example β The standard deviation of closing prices of two shares X and Y were Rs. 5 and Rs. 50, and mean closing prices during a week were Rs. 10 and Rs. 1000 respectively.
If we look only at standard deviation, we might invest in share X because it has less volatility. But using CV:
( CV_x = \frac{5}{10} \times 100 = 50% )
( CV_y = \frac{50}{1000} \times 100 = 5% )
Now we change the decision in favor of Y, as fluctuation in Y's prices is much lesser than share X. The coefficient of variation is a good measure for comparing riskiness in this case.
β Key Takeaways
The most critical concepts from this lecture are: (1) Range is the simplest measure of dispersion but only uses two extremes, while Variance and Standard Deviation measure the average squared deviation from the mean and its square root. (2) The standard deviation for a sample is computed by dividing the sum of squared deviations by ( n-1 ), not ( n ). (3) For normally distributed data, approximately 68%, 95%, and 99.7% of values fall within Β±1, Β±2, and Β±3 standard deviations of the mean, respectively. (4) Coefficient of Variation is essential when comparing variability between datasets with different units or vastly different means (e.g., comparing stock price risk). (5) Businesses use dispersion measures to assess risk, consistency, and to make data-driven decisions beyond just averages.
π§ Quick Revision Questions
- What is the difference between variance and standard deviation?
- In the ABC Real Estate example, what was the mean and standard deviation of the time to sell homes?
- According to the rules of standard deviation, what percentage of the population falls within Β±2 standard deviations of the mean?
- Why is the coefficient of variation preferred over the standard deviation when comparing the riskiness of two stocks with very different average prices?
- In the bank waiting time example, how did the manager justify his statement that customers would "almost certainly" wait not more than six minutes?
π Lecture 28 β Charts and Graphs
π Overview: This lecture introduces the use of visual representations in marketing research data analysis, focusing on the three most common types of charts: pie charts, line charts, and bar charts. It matters because charts and graphs allow researchers to present complex data in an easily digestible and interpretable format at a glance.
ποΈ Topics Covered
The lecture begins by stating that data can be shown by charts and graphs in addition to text and tables. It then covers three main types of charts: the Pie Chart, which is used for representing categorical data as parts of a whole; the Line Chart (or Graph), which is used to show dynamic changes over time; and the Bar Chart, which uses bars to compare the magnitudes of different categories. Variations of the bar chart, such as using pictures instead of bars and the grouped bar chart, are also discussed.
π Lecture Summary
Charts and Graphs
Data can be shown by charts and graphs in addition to the text and tables. Although the text and tables are useful for explaining and interpretation, charts and graphic illustration may add to the value of information because they provide information to the reader in a glance. Due to availability of computer software, it is now very easy and quick to make these charts and graphs. There are many kinds of charts that can be used in data analysis, but we will mention only three types which are more common: Pie chart, Line chart (graph), Bar chart.
Pie Chart
A pie chart is probably the most familiar chart used in representing quantitative data. A pie chart is simply a circle divided into sections, with each of the sections representing a portion of the total. As the sections are presented as part of the whole, pie chart is particularly more useful in showing relative size. Pie chart is prepared for categorical data.
π Definition β Pie Chart: A circle divided into sections, with each section representing a portion of the total. π Formula: The pie is based on 360 degrees, and each sliceβs angle is calculated based on its percentage of the whole. π Example: A pie chart showing the share of market of different brands of mobile phone sets, where Brand A has 35%, Brand B has 31%, Brand C has 20%, and Brand D has 14%.
Pie chart is based on the fact that the circle has 360 degrees. The pie is divided into different slices according to percentage in the category. Different colors can be used. To show emphasis on some category of data, the relevant portion of the pie can be lifted (it called exploding the pie). A pie chart can have many sections or slices, but it is recommended that no more than six sections should be generated in a pie.
π Example: In order to determine the significant source of business, WR hotel examined the check-in cards of 1000 customers randomly. They found the following break-up of data: Individual travelers 230, Tour groups 125, Business travelers 378, Government officials 143, Others 124. The solution involves calculating relative frequencies: Individual travelers (23.0%), Tour groups (12.5%), Business travelers (37.8%), Government officials (14.3%), Others (12.4%). π‘ Why this matters: This allows the hotel to see at a glance that Business travelers are their most significant customer source.
Bar Chart
In a bar chart, each category is depicted by a bar, vertical or horizontal. The length or height of the bar shows the frequency or percentage of observations falling into a category. The lengths or heights of different bars allow the user to compare the magnitudes of different categories easily.
π Example: The expansion of a bank in terms of opening of new branches from 2001 to 2007 was as follows: 2001 β 15, 2002 β 28, 2003 β 14, 2004 β 21, 2005 β 19, 2006 β 23, 2007 β 25. This data can be represented on a bar chart with years on the X-axis and the number of branches on the Y-axis.
Variation in Bar Chart
The bar chart has variations. Pictures can be used instead of bars, e.g., people for population, pictures of cars for automobile production, and piles of 1000 rupees note for sales. Another variation of bar chart used frequently is the grouped bar chart where more than one category can be captured and can be compared side by side in different colours.
π Example: A survey of retail sales of split air conditioners of 3 brands S, T, Y from 2000 to 2007, revealed the following values in million of rupees: Year 2000: S=4.3, T=3.2, Y=5.2 Year 2001: S=4.8, T=4.2, Y=6 Year 2002: S=6, T=5, Y=6.2 Year 2003: S=5.8, T=3.1, Y=6.8 Year 2004: S=7.2, T=3.7, Y=7.5 Year 2005: S=4.6, T=4.1, Y=6 Year 2006: S=6, T=5.1, Y=7 Year 2007: S=7.5, T=5.4, Y=8.5
A line chart with three different lines for S, T, and Y can be used, or a grouped bar chart with three bars side-by-side for each year. π‘ Why this matters: These variations allow for more engaging and comparative visualizations of data.
Line Chart or Graph
The pie chart is a one-scale chart. It is best used for static comparison, that is, the phenomenon at one time. The bar chart is having two dimensions, one of which usually is the time. It shows dynamic relation of the changes with time such as time series fluctuations. In the chart, the X-axis represents time and the Y-axis, the values of the variables. More than one variable can be plotted on the same graph but each variable is represented by different lines in different colors or form (dashes or dots) with explanations in the legend at the bottom of the graph.
π Example: Using the same data for brands S, T, and Y from the bar chart variation example, a line chart shows the trend of sales for each brand over the years 2000-2007.
β Key Takeaways
For exam success, a student must remember that pie charts are best for showing parts of a whole (relative size) with categorical data, and it is recommended to have no more than six slices. Line charts are specifically used to show trends and dynamic changes over time, with time on the X-axis. Bar charts are for comparing magnitudes across categories, and they have variations like using pictures or creating grouped bar charts to compare multiple variables side-by-side. Finally, computer software makes the creation of these charts quick and easy, but the analyst must choose the correct chart type for the data and the message.
π§ Quick Revision Questions
- What is the primary purpose of a pie chart, and what type of data is it best suited for?
- When would you choose a line chart over a bar chart to represent data?
- What is a "grouped bar chart" and what advantage does it offer over a standard bar chart?
- How is a pie chart's "exploding" feature used to improve data presentation?
- What is the recommended maximum number of sections (slices) a pie chart should have, and why?
π Lecture 29 β Hypothesis Testing
π Overview: This lecture introduces hypothesis testing as a key component of data analysis in marketing research. It covers the formulation of null and alternative hypotheses, the distinction between one-tailed and two-tailed tests, and the fundamental steps involved in conducting a statistical hypothesis test.
ποΈ Topics Covered
This lecture begins by defining a hypothesis as an educated guess about relationships between variables, which must be empirically tested and formally stated. It then explains the null and alternative forms of hypothesis, followed by the concepts of one-tailed and two-tailed tests with specific examples related to marketing metrics like defective parts and sales returns. Finally, it outlines the step-by-step process of hypothesis testing and mentions the selection of appropriate statistical techniques based on research objectives.
π Lecture Summary
Data analysis
Hypothesis is an educated guess, a tentative statement about the relationship between two or more variables. It is to be empirically tested and be stated before the marketing project begins. Hypotheses must be formally stated, as they are focal points for researchers in marketing. Hypotheses may be in operational (general) terms or null and alternative forms.
Null and Alternative Form of Hypothesis
Testing of hypothesis usually begins with stating the hypothesis in a null and alternative form. For example, we might want to see whether mean age of a class of consumer is 30 years. π Definition β Null hypothesis (Hβ): A statement that there is no effect or no difference, serving as the default assumption to be tested. It is written as Ho: ΞΌ = 30. π Definition β Alternative hypothesis (Hβ): A statement that contradicts the null hypothesis, indicating the presence of an effect or difference. It can be written as Hβ: ΞΌ β 30 or Hβ: ΞΌ > 30. π Formula: Ho: ΞΌ = 30 and Hβ: ΞΌ β 30 β The null states the mean is exactly 30, while the alternative states the mean is not equal to 30. π Example: For a consumer class mean age, the null hypothesis is Ho: ΞΌ = 30, and the two-tailed alternative is Hβ: ΞΌ β 30.
One tailed and two tailed test
The previous example of alternative hypothesis is two-tailed as we will reject the null hypothesis if the mean age was lesser or greater than 30. π Definition β One-tailed test: A test where the alternative hypothesis specifies a direction (greater than or less than a value). Another alternative hypothesis in this situation could be that the mean age is greater than 30. In this case we would write null and alternate form of hypothesis as: Ho: ΞΌ = 30 Hβ: ΞΌ > 30 Here the hypothesis is one-tailed as we have a specific direction in mind for the alternative hypothesis. We can also phrase the null hypothesis to cover a range of values. For example, Ho: ΞΌ β€ 30 Which implies an alternative hypothesis Hβ: ΞΌ > 30 Here again one-tailed test would apply. The researcher should be careful to phrase the alternative hypothesis in a way as to accept the alternative hypothesis that is of real interest if null hypothesis is rejected.
One-Tailed or Two-Tailed Test
- Defective parts are more than 2% β One-tailed
- Sales returns are less than 4% p.m. β One-tailed
- Within one per cent of the mean β Two-tailed
- Between 5% and 6% β Two-tailed We use one and two-tailed concept when we look critical values in the probability tables. We should have this concept clear so that we can look into the appropriate table.
Steps in Hypothesis Testing
Following steps are usually followed in hypothesis testing:
- Formulate a null and alternative hypothesis
- Specify the significance level
- Select the appropriate statistical technique according to the nature and type of data collected
- Perform the statistical test applying the technique above
- Look for the value of test statistics (critical value) in the relevant standard normal table on the confidence level as specified in step #2 above
- Compare the value of statistics as calculated in step #4 with critical value and accept or reject the hypothesis
- At the end researcher draws conclusion.
Different Statistical Tests
Which statistical technique or test to select for our analysis or hypothesis testing depends on our objectives of research project. Objectives may be translated into research questions and/or research hypotheses. There are many statistical tests and techniques which are used in data analysis in marketing research. Some of them are described in the next lectures in detail. π‘ Why this matters: Choosing the wrong statistical test can lead to incorrect conclusions, making it essential to align the test with the research objectives and data type.
β Key Takeaways
Hypothesis testing begins with formally stating a null hypothesis (Hβ) and an alternative hypothesis (Hβ), which must be empirically tested. The key distinction between one-tailed and two-tailed tests depends on whether the alternative hypothesis specifies a direction (e.g., greater than) or simply a difference. The process involves seven steps: formulating hypotheses, setting a significance level, selecting and performing a statistical test, comparing the calculated test statistic with a critical value, and drawing a conclusion. The choice between one-tailed and two-tailed tests is critical when determining critical values from probability tables.
π§ Quick Revision Questions
- What is the difference between a null hypothesis and an alternative hypothesis?
- When would you use a one-tailed test instead of a two-tailed test?
- What are the seven steps in hypothesis testing?
- In the example "Sales returns are less than 4% p.m.", is this a one-tailed or two-tailed test?
- Why must a hypothesis be stated before the marketing research project begins?