There is no magic formula to determine the proper sample size – it depends on the complexity of your research, how homogenous the population is, and time and human resources you have available to compile and analyze data. Descriptive Statistics Once you have performed your research and gathered data, you need to perform
Evidence Based Library and Information Practice 2007, 2:1
36
Table 1. Examples of hypotheses. data analysis. Choosing the appropriate statistical method for the data is crucial. The bad news is, this means you have to know a
whole lot about your data – is it nominal, ordinal or ratio? Is it normally distributed? Let’s start from the very beginning.
A clear understanding of librarians’ perceptions of EBP is necessary to inform the development of systems to support EBP in librarianship. The following research questions were posed:
1. What are the perceptions of librarians of EBP? 2. Does institution type the librarian works at affect perception? 3. Does length of service of the librarian affect perception?
What are the hypotheses? There are three being provided. Here is a rephrasing of number 3:
H0 = “Length of service of librarians has no affect on the perception of EBP” H1 = “Length of service of librarians affects the perception of EBP”
What are the Type I & II error possibilities?
The real situation (in the population) H0 is true H1 is true
No error
Type II error
Result of Research (from sample):
H0 is proven (length of service doesn’t affect perception) H1 is proven (length of service does affect perception)
Type I error
No error
What are the dependent and independent variables? The researchers are attempting to determine whether length of service can predict perception of EBP, or to rephrase, is perception of EBP dependant on length of service. Therefore:
Dependent variables: perception of EBP Independent variable: length of service
Evidence Based Library and Information Practice 2007, 2:1
37
Levels of Measurement Nominal variables are measured at the most basic level. They are discrete levels of measurement where a number represents a category (i.e., 1 = male; 2 = female), but these numbers do not imply order and mathematical calculations cannot be performed on them. You could just as easily say, 1 = male and 36,000 = female ‐ this doesn’t mean that females are 35, 999 times bigger or better than males! Nominal variables are of the least use statistically. Ordinal variables are also discrete categories, but there is an order to the categories; they increase and decrease at regular intervals. A
good example is a Likert scale: 1 = very poor; 2 = poor; 3 = average, etc. In this example, you can state 1 is ‘less’ or ‘smaller’ or ‘worse’ than 2. The disadvantage of ordinal variables is that you cannot measure in between the values. You do not know how much worse 1 is than 2. Ratio (sometimes known as scale, continuous or interval) variables are the most robust, statistically, of variable types. Ratio variables have natural order, and the distance between the points in the same. Think of pounds on a scale. You know that
Table 2. Examples of sampling.
The sampling frame was the database of all librarians (defined as those who hold an MLS) who were members of the Canadian Library Association in March 2005. A total of 5,683 librarians were on the list. The list was divided up by type of library worked at (academic, public, school, special, and other / not stated). A proportional random sample of 210 was then selected. This ensured that even at a return rate of 40% a final sample size of 150 would be achieved. Is this a random sample? On first glance, yes. However, this is only a true random sample if all librarians in Canada belonged to the Canadian Library Association. The design of this study means that the results can only be generalized to Canadian Library Association members, not to Canadian librarians. What sampling technique is used? This survey used stratified random sampling to ensure that all types of librarians would be represented, as illustrated in the chart below. Please remember that all values in this table are for demonstration purposes and do not accurately reflect reality. Academic
Librarians Public Librarians
School Librarians
Special Librarians
Other / Not Stated
Totals
Real Proportion
1136 (20%) 2273 (40%) 568 (10%) 582 (15%)
582 (15%) 5683
Sample Size 42 (20%)
84 (40%)
21 (10%)
31 (15%)
31 (15%)
210
Evidence Based Library and Information Practice 2007, 2:1
38
100 is lighter than 101. You also know that 101 is 1 pound heavier than 100. Finally the scale is continuous; it is possible to weigh 100.58 pounds. The power of the ratio variable is important to keep in mind for your study. For example, rather than asking subjects to tick off an age category in a box, you can ask them to fill in their age. This gives you the freedom to keep it as a ratio variable, or to round the ages up into
appropriate ordinal values. Measures of Central Tendency The theory of normal distribution tells us that, if you tested an entire population, the result (parameter) would look like a bell curve, with the majority of values grouped in the middle. A good example of this would be scores on test.
Table 3. Examples of variables.
Selection of variables used in the study
Variable Name Variable Label Values
TYPE
Type of library worked at 1 = academic, 2 = public…
LENGTH
Length of service
INCOME
Income of respondent 1 = under 30,000, 2 = 31,000‐ 40,000…
AGE Age of respondent
EBP_AWARE Answer to the question I have heard of EBP
1 = yes, 2 = no
EBP_SCORE Score on the EBP Perceptions Test
What level of measurement is TYPE? TYPE is a nominal measurement. The numbers represent types of libraries, but no mathematical calculations can be performed on them. EBP_AWARE is also a nominal measurement. What level of measurement is LENGTH? Because there are no values set for LENGTH it is a ratio variable. Each librarian’s length of service will be entered in years. EBP_SCORE and AGE are also ratio variables. What level of measurement is INCOME? INCOME is an ordinal variable. It has numbers representing categories, but there is a clear ranking. Librarians in category one earn less money than librarians in category two.
Evidence Based Library and Information Practice 2007, 2:1
