Tuesday, March 4, 2014
Sampling Strategy and Sample Size for a Quantitative Research Plan
The purpose of the sample and the strategy behind the sample design is to ensure a study’s generalizability (Frankfort-Nachmias & Nachmias, 2008). An important aspect of research is its ability to be generalized to wider populations. When the inferences and interpretations derived from a sample are not generalizeable, the research is of little scientific value. When a sample is adequately representative of the population under scrutiny, inferences made about the sample can be applied or implemented in the population (Frankfort-Nachmias & Nachmias, 2008). The purpose of this paper is to recommend a sampling strategy and sample size for my research plan as well as a rationale for that recommendation. In addition it will describe the population and its size.
The population used for this study is adults of both genders, between the ages of 30 and 65, who have been diagnosed with a rare type of cancer at least six months prior to becoming a participant in this study. Participants cannot have been diagnosed with a major depressive disorder or depressive symptoms lasting more than six months prior to being diagnosed with cancer. Participants will have participated in either a face-to-face support group or an online, disease-specific support group for at least three months, but no more than nine months. Participants in each group are not, to their knowledge, in end-stage cancer, which, for the purposes of this study, is defined as having approximately six months or less to live.
Although there is no general consensus of how many individuals are living with rare types of cancer, the Rare Diseases Act of 2002 (2002) defines rare diseases as those that affect fewer than 200,000 people each year in the United States. For this study, rare cancer is defined as cancers that make up less than 2% of all cancer diagnoses. Because of the lack of information available on the number of individuals living with cancers that fit this description, I have estimated the population at 200,000 individuals.
My study is a quasi-experimental design, which calls for the utilization of two groups; one whose members have participated in face-to-face support for between three and nine months, and a second group, whose members have participated in online, disease-specific support forums for the same amount of time. My sampling strategy will be to obtain a probability sample. Probability sampling is based on probability theory which increases the chance that each member of the population will have an equal chance of being included in the sample. The probability sample is obtained randomly and reduces the chance of a non-representative sample (Frankfort-Nachmias & Nachmias, 2008).
After eliminating individuals who have not been diagnosed for at least six months, are in late stages of cancer, or had been diagnosed with a major depressive disorder or with depressive symptoms lasting more than six months, prior to being diagnosed with cancer, I will utilize systematic sampling. To implement systematic sampling, I will determine an interval and select every nth response according to the time and date the individual's response was received. The value n is the integer value of the ratio of the size of the population to the size of the sample. For example, if I receive 400 responses and I intend to use 34 participants for my sample, I would take the number of responses and divide it by the number of participants I need for my sample, which, in this case is 400/34, which equals 11.76, or approximately 11. I would select every 11th person, after randomly choosing the starting point. This is called a 1-in-11 systematic sample. This selection method will decrease threats to validity that could be caused by selection bias.
Selection for Each Group
To obtain the sample group that participates in face-to-face support, I will advertise for participation in various local cancer support groups. For the group that participates in online, disease specific support forums, I will post a call for participation in at least five disease-specific support forums provided by the Association of Cancer Online Resources (ACOR). ACOR is an online cancer community that provides support group forums for over 80 types of cancer as well as a forum for rare cancers. Responses to these notices will be systematically selected as previously described. Selection will be made from the number of responses, after excluding patients who do not fit the description of the population as defined.
Choice of Sample Size
Because of the lack of consensus on the exact population of people living with rare types of cancer, and because of the nature of my research and few representative studies, I decided to utilize Cohen's d and a t test for two independent samples. Using a power or .80 and a medium to small effect size of .70 (described as between .50 and .80), the necessary sample size should include 34 participants in each group. Using an inversion of Fort-Nachmias and Nachmias' (2008) formula (SE= s/√n), n=s2/SE2 I obtained a recommendation of 34 individuals for each group. Using G Power software program, the recommendation was for 34 participants in each group as well.
I found two studies that related somewhat to my study. Applying the formula for Cohen's d, I subtract the means (5.6-3.5 = 2.1) and average the standard deviations (3.5 + 2.7/2 = 3.1). Then I divided the means by the average of the standard deviations (2.1/3.1 = .68). Using .68 or approximately .70 as an average effect size, and the tables provided by Burkholder (n.d.), I found that for comparing two groups, I need two samples with approximately 34 individuals in each group. Since all methods of calculation agree that the most appropriate sample size is 34 individuals for each group, I am confident that by utilizing this number of participants, I can ensure my samples are an adequate size.
Burkholder, G. (n. d.). Sample size analysis for quantitative studies. Retrieved from https://class.waldenu.edu/bbcswebdav/institution/USW1/201430_01/XX_RSC H/RSCH_8200/Week%206/Resources/Resources/embedded/Sample_Size_.pdf
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, 175-191.
Frankfort-Nachmias & Nachmias (2008). Research Methods in the Social Sciences (7th ed.); Worth Publishers: NY
Rare diseases act of 2002. 280, 107th Cong., U.S. G.P.O. 116 (2002) (enacted).