Methods of Communication Research and Statistics Online Workbook
|
SPSS exercise 2.7
In the variable 'ChurchGoing' (dem84), in the WatchAndListenSurvey.sav database, measurements are taken of how often people go to church. When you calculate a measure of central tendency for this variable, it gives information about all the respondents, including those who answered 'no answer' or 'not applicable'. Before doing the exercises below, you first have to list these options as 'missing values' in SPSS.
a. What percentage of the respondents goes to church once a month or less?
b. What is the level of measurement of this variable?
c. What conclusion can you draw? When giving your answer, you must include the following: the 'content' of the variable, the units of analysis, the correct number for the measure of central tendency (round off to two decimals), and the interpretation of the measure of central tendency.
Hints
1 Determining the level of measurement of a variable
Nominal scale
In a nominal scale, the characteristics are assigned a random value. The sequence of the characteristics does not matter, as can be seen on the scale below. Moreover, the distance between the letters themselves has no real meaning.
3 1 4 5 2
Ordinal scale
In an ordinal scale, the characteristics are not assigned a random value, but the scale shows their rankings. This can be seen on the scale below. The values are in alphabetical order. The distance between the letters remains random, however, and without any real meaning.
1 2 3 4 5
Interval scale
In an interval level of measurement, the scale indicates a sequence. The distance between the letters themselves also has real meaning. For example, the difference between '2' - '1' = '4' - '3' and the distance '3' - '2' = '5' - '4'. However, there is no zero point. An example of an interval scale is the temperature in degrees Celsius. The difference in temperature between 5 and 10 degrees Celsius is the same as that between 35 and 40 degrees. However, the zero point is random - in other words, 0 degrees Celsius is not the lowest possible temperature. This means that you cannot say that 40 degrees Celsius is four times greater than 10 degrees Celsius, because the relationships between two numbers are not the same (because of the lack of identical distances, and a natural zero point).
1 2 3 4 5
Ratio scale
The ratio scale has all the same characteristics of an interval scale. But it also has a natural zero point (0). This means that the differences between the numbers on the scale have a real and identical meaning, as does the relationship between two numbers. Therefore, you can say that a person who has four children has twice as many children as someone who has two.
0 1 2 3 4 5
2 Measures of central tendency and measures of dispersion selection chart
|
Measure of central tendency |
Mode |
Median |
Mean |
Measure of dispersion |
|
Interquartile range |
Standard deviation, Variance |
1. Select |
1.1 Specifics |
What occurs most frequently? |
What is the median value? |
What is the balance point of the distribution? |
1.2 Minimum level of measurement |
Nominal |
Ordinal |
Interval |
1.3 Shape of the distribution |
Usable with every distribution shape |
Usable with every distribution shape |
Not optimal for very skewed distributions |
2. Calculate |
Analyze-> Descriptive Statistics -> Frequencies -> Statistics |
Analyze-> Descriptive Statistics -> Frequencies -> Statistics |
Analyze-> Descriptive Statistics -> Frequencies -> Statistics |
3. Check |
Unusable with a distribution with many peaks, e.g. in the case of an ungrouped interval variable. |
|
The greater the difference between the mode, median, and the mean, the more skewed the distribution. |
4. Conclude |
General:
1. Give your reasons for selecting the measure of central tendency and measure of dispersion (see Select and Check).
2. Use the value of the measure of dispersion to comment on the accuracy of the calculated measure of central tendency: the smaller the distribution, the better the centre typifies the distribution. |
Mode: "The most common value of variable X is A.". |
Median: "The median value of variable X is A" or "Half of the units of analysis have A or less on variable X, and half of the units of analysis have A or more on variable X."
Interquartile range: "The mean deviation from variable X compared to the median is A" or "The middle half of the observations lies between (first quartile limit) and (third quartile limit).
|
Mean: "The units have an average A on variable X."
Standard deviation, Variance "The mean deviation in relation to the mean."
|
|