In our two recent initiatives on Entertainment-Education — Farm IPM Radio supported by the Rockefeller Foundation and Environmental Radio Soap Opera funded by the World Bank Development Marketplace 2005 Award — we took great pains in analyzing the audience. Research-based information on target audience characteristics, needs and preferences was to used to enhance the values grid and design the drama content to ensure that the soap operas resonate with the audience.

In an earlier post, we talked about the what, why and how of audience research. Supposing you have already gathered your audience analysis data, the next step is to review each questionnaire to make sure that the respondent has not skipped any question in the instrument. Once the completed questionnaires have been edited, the data need to be analyzed. This analysis phase can be relatively simple – such as manually determining the % of respondents giving specific answers or listing the various ways in which farmers said they might utilize a new practice. Whether your survey is simple or complex, it is best that the data are encoded, processed and analyzed using a statistical package.

The first step in data analysis is to code the data. Coding is the term used to describe the translation of question responses and respondent information to specific categories for analysis. The first stage of coding information involves the construction of a code book. A code book is a set of rules used to classify observations of variables into values that are transformed into numbers.

Sample Codebook

Q No.

Column

Variable Name

Codes

A

ID

Enter actual number

1

B

District

1 = Vi Thuy (1- 304)

2= O Mon (305- 606)

3= Chau Thanh A (607-908)

2

C

Sex

1= Male

2= Female

3

D

Source

1= Bought from seed grower

2= From own farm

3= Obtained from other farmers

4= Government extension office

5= Other (specify)

Variable naming rules
1. Keep column or variable names short; statistical packages such as SPSS won’t read variable names longer than 8 characters; variable names can have a maximum of 8 characters.
2. Must begin with a letter.
3. Cannot include a period, blanks or certain characters (!, ?, `, *)
4. Must be unique – this means that not two variables must have the same name.
5. Cannot include these reserved words (ALL, NE, EQ, LE, LET, BY, OR, GT, AND, NOT, GE, WITH)
6. Use one word – alphabet and numbers – but no space in between characters.
7. Variable names are not case sensitive, i.e. can be written in upper or lower case

Data entry
1. Use numerical codes for all questionnaire data.
2. Enter numbers only in a spreadsheet.
3. Enter only 1 answer for each column.
4. Create another column for another answer.
5. Remember that each row or line of the table represents one respondent or case, and each column represents one variable

ID

Sex

Age

Area

Variety

1

1

23

1.2

1

2

1

26

0.3

3

3

2

35

0.5

1

4

1

54

2.1

5

5

2

63

0.2

4

6

2

49

1.7


Data checking

1. After the data have been entered, check the file for wild codes and extreme values.

ID

Sex

Age

Area

Variety

1

11

23

1.2

1

2

1

26

0.3

3

3

5

35

0.5

1

4

1

54

2.1

5

5

2

63

0.2

4

6

22

49

1.7

1

In the sample data above, 11, 5, and 22 are considered wild codes for sex which generally has only two codes: 1=male, 2=female.

2. Some statistical packages enable one to check the data for outliers.

Data analysis
1. Get descriptive statistics (mean, median, mode, and range) for each numeric variable. Compare answers with earlier surveys – if possible. Are means too high or too low because of errors or extreme values from certain respondents?

2.Get frequency distributions for all variables. Compare results for different items in the set.

3. Analyze several variables at a time by looking at a group of related variables.
- Run crosstabs for pairs of variables you have selected.
- For nominal and ordinal variables, run a chi-square test and check if the relationship is significant.

Levels of measurement
The statistical measures to use with a variable depends on which type of scale it was measured on: nominal, ordinal, interval, or ratio. Nominal variables are numbers or words that are used merely as labels such as sex, district, and province. Ordinal variables are numbers or words that can be rank-ordered such as rating scales. Interval scale are numbers that have equal and consistent intervals but no true zero point, such as number of insecticide sprays. Ratio scale are real numbers that have a true zero point such as age, height, weight, and yield.

References
List, Dennis. 2005. Know your audience: a practical guide to media research. Original Books, Wellington, New Zealand.

Nachmias, C. F. and D. Nachmias. 1996. Research methods in the social sciences. St. Martin’s Press, New York.