Statistics: Assessing T-tests

To clearly identify what a t-test accomplishes in descriptive statistics it is imperative to understand what a t-test represents. A “t-test is a parametric statistical test for comparing the means of two independent samples” (Plichta & Kelvin, 2013, p. 464). Gosset developed the t-test for use in quality control at the Guinness Brewery and published his works under the pen name “Student” (Plichta & Kelvin, 2013). T-tests use assumptions related to the underlying variable of study, where it is assumed that the underlying variable is distributed normally (Fagerland, 2012, p.76).

In essence, the t-test examines differences between two groups on a variable of interest. Research Study Review Bradley (2012), proposed to study the effects of nurse practitioners (NP’s) utilizing a skin cancer screening tool along with receiving further education on skin cancer assessment and diagnosis. The study was predicated on the Health Promotion Model (HPM), which was selected based on it relevance to the practice of NP’s and their duty to promote healthy behaviors (Bradley, 2012).

“Evidence-based research showcased within the HPM can aid NPs in making practical recommendations to reduce skin cancer risk, specifically melanoma” (Bradley, 2012, p. 83). Study participants (NP’s) were provided with training and educational programs to increase their knowledge and subsequent assessment skills and documentation of physical skin assessments for patients. “The HPM was used in this study to examine the health status and health promotion behaviors of young adults with respect to screening for melanoma” (Bradley, 2012, p.83).

The study was conducted in two phases where “phase one utilized a quasi experimental design to examine characteristics of a single sample of NPs exploring the aspects of skin cancer screening phenomena among them in a single college health center setting” (Bradley, 2012, p. 83). NP’s were given a pre and post test to ascertain their cognitive processes and the changes in the ability to document properly based on the education they received (Bradley, 2012).

“In quasi-experiments, the researcher does not have the ability to randomly assign the participants or ensure that the sample selected is as homogeneous as desirable” (Levy & Ellis, 2011, p. 155). NP’s attitudes were also evaluated after receiving the educational component to determine if attitude changes were elicited (Bradley, 2012). The study revealed an increase of correct documentation of skin cancer screening and education related to skin cancer (Bradley, 2012). Study Variables and Statistics

The study utilized a convenience sample of six certified NP’s who ranged in age from 40-64 (Bradley, 2012). Convenience sampling chooses samples that are readily available, this type of sampling can possess bias due to a lack of a statistical method of choosing the sample (DiCalogero, n. d. ). The NP’s were given a Didactic skin cancer pre and post test, which consisted of 16 questions using a multiple choice format used to evaluate cancer knowledge and recognition (Bradley, 2012). “Test-retest reliability was r = .

90, no alpha was given” (Bradley, 2012, p. 84). A program evaluation questionnaire using a Likert Scale was given following the post test to assess the NP’s attitudes, knowledge and perceptions related to relaying the skin cancer data and future participation in skin cancer screening for the public using a five point scale (Bradley, 2012). NP’s were also asked to use the screening tool for documentation via a pictorial display to document lesions and note patient education given (Bradley, 2012). Data collection transpired over a 30-day period.

“Prior to the education, 30 randomly selected de-identified charts were retrospectively reviewed for skin cancer screening documentation” (Bradley, 2012, p. 85). At the conclusion of the study 22 additional charts were audited to note if improvement in the skin assessment, education, and referrals, if needed were charted (Bradley, 2012). The independent variable in this study is the six certified NP’s attitudes and acceptance. An independent variable is the variable you have control over, what you can choose and manipulate”, (North Carolina State University [NCSU], n.

d. , para. 1). The dependent variable is the increase in skin cancer education knowledge. Dependent variables “measure the effect of some other variable”, (Plichta & Kelvin, 2013, p. 457). Other variables are presence of chart documentation materials on skin cancer. A paired sample t-test evaluated the difference in the NP’s correct responses between the pre test education and post test education responses on the Didactic skin cancer test. The “mean=33. 6458, standard deviation (s)=85.

7564, number of questions n=16, t-statistic (t)(15)=1. 57 at a level of significance alpha (? )=0. 10, the upper critical value t=1. 34, which suggests there is enough statistical evidence to suggest that the Didactic increased the mean of correct answers to questions” (Bradley, 2012, p. 85). “The critical value of the t-statistic is the threshold to which the computed value of the t-test is compared to determine whether the null hypothesis is rejected” (Plichta & Kelvin, 2013, p. 98).

The t-test offers validation that the NP’s have increased knowledge of skin cancer when the post test is compared to the pre test, due to an increase in the mean percentage (9. 33%) of correct answers. A two sample t-test was used to evaluate the use of the documentation tool to assess for skin cancer. This used the mean percentage difference in the pre-education and post education chart documentation (Bradley, 2012). The chart audit “revealed improvement of 223. 4% in the proper documentation of skin cancer findings and education” (Bradley, 2012, p.85).

This used discrete variables of “yes and “no” to indicate if the chart documentation tool was used to calculate a mean difference between pre and post test. “The two sample t-test tests whether or not two independent samples have different mean values on some measure” (Kaplan University, n. d. , p. 1). The NP program evaluation used a Likert scale tool and found that “participants felt that the training was extremely relevant to their practice with an increased willingness to suggest the training to their peers” (Bradley, 2012, p. 87) .

The test questions were identified by their mean response using standard deviation, and what the minimum and maximum ranking was on the one to five scale for each question. Statistical Assumptions T-tests require that the outcome variable is continuous and that two groups are being compared (Heavey, 2011). The variable of education which was tested using the Didactic screening tool and the program evaluation assessment of attitudes are continuous variables which can “take on any possible value within a range” (Plichta & Kelvin, 2013, p. 456).

It is also necessary that “only two measurements (pre and post test) on the same person are taken and a matched control subject are compared, the total sample size, whether the two measures of the variable are normally distributed and the measurement scale of the variable measuring the characteristic of interest” (Plichta & Kelvin, 2013, p. 130). This study used two measurements pre and post test education levels of NP’s on skin cancer which were conducted on the same six NP’s, both pre and post test. The use of a control subject was completed in terms of the 30 pre test chart audits and the 22 post test chart audits of documentation.

The sample size of six was small and certainly could be limiting with regard to the assumptions. However, “the paired t-test can be used with confidence, and there is a relatively low risk of error if just a few assumptions are violated” (Plichta & Kelvin, 2013, p. 130). The t-test illustrates an increase in the mean percentage of questions that are correct between the pre and post test, validating the hypothesis that NP’s have increased knowledge of skin cancer after receiving education and chart documentation tools were utilized more as a result (Bradley, 2012). Levels of

Measurement

The multiple choice Didactic cancer screening test would be an example of the interval scale. “The most common examples of interval scales are scores obtained using objective tests such as multiple-choice tests of achievement” (University of West Florida, n. d. , para. 5). The pre test and post test chart audit used nominal terms of “yes” or “no” to determine the frequency number percent change of charts pre and post test that utilized the documentation education tool. Nominal measures “are the lowest level of measurement that organizes data into discrete units” (Plichta & Kelvin, 2013, p.

460). The intervention program evaluation utilized a Likert scale which is an ordinal measurement. “Ordinal measurement scales rank participants on some variables such as measurements of subjective states” (Plichta & Kelvin, 2013, p. 460). The paired t-test used the interval scale of measurement, the two sample t-test utilized the nominal measurement scale and the evaluation of attitudes used the ordinal scale. “The two sample t-test tests whether or not two independent samples have different mean values on some measure”, (Kaplan University, n. d. , p. 1).

The two sample t-test used nominal variables that were dichotomous, and illustrated their mean frequency change from pre test to post test, making the use of this test for the chart audit appropriate. The paired sample t-test should be assessed on an interval-or ratio-level of measurement where the predictor variable should be a nominal-level that includes just two categories” (O’Rourke, Hatcher, & Stepanksi, 2005, Chapter 8), which was appropriate in this case and used the difference in means between the pre test and post test responses of the NP’s. Data Displays The study used four data displays that were easy to understand.

The first table showed pre test and post test numbers of correct responses to the Didactic screening tool of 16 questions. It demonstrated the absolute change and the percent change in the six NP’s scores. Table two analyzed chart documentation per the 30 charts involved in the pre test audit using “yes” and “no” responses to questions and frequency of answers. Table three analyzed the documentation in the post test audit using “yes” and “no” responses and frequency of answers. Table four illustrates the mean score for the intervention program evaluation tool that the NP’s filled out using the Likert Scale tool.

It also illustrates standard deviation and minimum and maximum ratings on each question. All of the data displays were easy to understand, outlined the intent of the display, and defined the information clearly in each display. In conclusion this study appears to use t-tests and paired t-tests appropriately. The implementation of the education on skin cancer yielded higher post test scores and a increase in the mean scores as compared to the pre test. This meets the criteria put forth in the textbook which states that the “t-test is a parametric statistical test for comparing the means of two independent samples” (Plichta & Kelvin, 2013, p.

464). The test provides valuable research for NP’s to utilize in the future for assessing and treating patients whom have developed skin cancer. References Bradley, H. B. (2012). The implementation of a skin cancer screening tool in a primary care stetting: A pilot study. Journal of the American Academy of Nurse Practitioners, 24(2), 82-88. http://dx. doi. org/10. 1111/j. 1745-7599. 2011. 00669. x DiCalogero, S. (n. d. ). Statistical Sampling: A guide for gathering data [PowerPoint slides]. Retrieved from Tidewater Community College: www. tcc. edu/vml/documents/StatisticalSampling. pptx_ Fagerland, M. W.

(2012). T-tests, non-parametric tests, and large studies–a paradox of statistical practice? BMC Medical Research Methodology, 12(1), 76-83. http://dx. doi. org/10. 1186/1471-2288-12-78 Heavey, E. (2011). Statistics for nursing. Sudbury: MA: Jones and Bartlett Learning. Kaplan University. (n. d. ). Overview. Retrieved from http://kucourses. com/re/DotNextLaunch. asp? courseid=8514503 Levy, Y. , & Ellis, T. J. (2011). A guide for novice researchers on experimental and quasi-experimental studies in information systems research. Interdisciplinary Journal of Information, Knowledge, and Management , 6, 151-161.

North Carolina State University. (n. d. ). Independent Variable. Retrieved from http://www. ncsu. edu/labwrite/po/independentvar. htm O’Rourke, N. , Hatcher, L. , & Stepanksi, E. (2005). A step by step approach to using SAS for univariate and multivariate statistics (2nd ed. ). Cary, NC: SAS Institute. Plichta, S. B. , & Kelvin, E. A. (2013). Munro’s statistical methods for health care research (6th ed. ). Philadelphia, PA: Wolters Kluwer Health/Lippincott, Williams & Wilkins. University of West Florida. (n. d. ). Scales of measurement. Retrieved from uwf. edu/pcl/research/edf6481/week04/files/ScalesofMeasurement. rtf