Data Analysis
The researcher used SPSS version 17 for data analysis (Pallant, 2005). This provided a good chance for the researcher to analyse qualitative data after coding. Data analysis would involve the use of both descriptive and logistic regression. Data for correlation between variables would mainly consist of descriptive statistics. For instance, correlation tests would show whether the use of agency nurses correlate with patient outcomes. It would show whether such “relationships are positive or negative” (Pallant, 2005). In addition, descriptive statistics would indicate the extent of such relationships.
Descriptive statistics shall provide fundamental elements of information collected for the research questions. In other words, we would obtain summaries and measures about data collected. For instance, the researcher will describe mean and median to show the central tendency and mode of data to identify the most common measurement in collected data. Such data are useful for making inference. The standard deviation, coefficient of variance and variance shall show any dispersion in analyzed data. Any calculated value would assist the researcher to understand deviation of values from the mean range. The researcher shall also know normality in the data through tests for skewness and Kurtosis. Most statistical techniques assume that distributions of scores on dependent variables are normal. In this case, a normal distribution represents a symmetrical, bell-shaped curve, which has the greatest frequency of scores in the middle, with smaller frequencies towards the extremes. Skewness tests for instrument normality indicate the tilt or lack of it in distribution of respondents. These are right and left skewness whereby right skew are common. Some authors have argued that skew should have a range of +2 to -2 to indicate normal distribution. However, others statisticians apply a stringent rule of + 1 to – 1 to assess normality of instruments. There are several statistical techniques that the researcher can use for making assumptions about the data collected.
The logistic regression would enable the researcher to explore how well a “set of variables explains the categorical dependent variable” (Pallant, 2005). The approach would allow the researcher to test for categorical outcomes in two or more categories. These variables would be either categorical or continuous, or they may consist of both outcomes. Logistic regression would show the adequacy of variables by exploring their suitability (Pallant, 2005). The approach would account for the relative significance of study variables and their association. Logistic regression would provide “a summary of the accuracy of the classification of cases based on the mode, allowing the calculation of the sensitivity and specificity of the model and the positive and negative predictive values” (Pallant, 2005, p. 163). This method of data analysis would not make any assumption about the scores of variables.
Techniques for data analysis
The researcher will use two techniques to analyze collected data, which would include descriptive statistics and logistic regression. The software for data analysis shall be IBM SPSS (Statistical Package for the Social Sciences) version 17.
The initial stage would involve screening data for any errors. This is necessary because of mistakes, which occur when entering data. Thus, the researcher would be cautious to avoid mistakes. He will focus on outliers in the data. Data screening would involve checking, identifying, and correcting errors (Pallant, 2005).
Descriptive statistics
Descriptive statistics would be useful for analyzing features of the collected data. The researcher shall present simple summaries regarding the use of agency nurses and patient outcomes related to failure to rescue, mortality, and morbidity. With other graphical elements, descriptive statistics would present a good basis for analysis of quantitative data in the study.
The researcher will use descriptive statistics to describe what results from the analysis depict. This would be data in a manageable form, which many people can comprehend. The researcher will present all relevant background information of agency nurses, such as age, education level, gender, economic status, total number of participants through descriptive statistics. These will be in both numerical and graphical approaches.
Descriptive statistics for the study would include “the mean, standard deviation, range of scores, skewness, and kurtosis” (Pallant, 2005, p. 49). For categorical variables, under descriptive statistics, the study shall use frequencies to indicate the number of agency nurses who gave specific responses for specific study issues. For instance, how many male or female agency nurses concur with negative or positive patient outcomes? The researcher shall not use descriptive statistics for categorical variables like gender of agency nurses.
Data shall also have continuous variables such as age of agency nurses. Descriptive statistics would allow the researcher to get the mean, median, and standard deviation of research participants.
Logistic regression
The researcher will use logistic regression to show the correlation between independent and dependent variables. Logistic regression would be suitable for analyzing categorical variables like Yes/No, Increase/Decrease questions, which will be dependent variables. Such question will aim to explain the correlation between the use of agency nurses and patient care outcomes in terms of mortality, morbidity, and failure to rescue.
Rationale for data analysis technique
After screening and cleaning collected data for errors, the researcher will start descriptive analysis. Descriptive statistics is important because it shall allow the researcher to describe characteristics of study variables (agency nurses and patient care outcomes). It shall also provide opportunities for the researcher to check for “any violation of the assumptions underlying the statistical techniques that he will use to address specific research questions” (Pallant, 2005, p. 49). Descriptive statistics shall use both numerical and graphical approaches. In numerical approach, the researcher would show variable mean, mode, standard deviations, and variances among others. They shall provide accurate and objective accounts of the use of agency nurses and patient care outcomes. The researcher will also use a graphical approach for background variables in which he may display in forms of charts, tables, and graphs. Graphical presentations provide simple and easy to understand results of data analysis. Therefore, descriptive statistics would allow the researcher to put data in their manageable form for people to understand. Data usually have many measures. He will also be able to present data in simple and sensible way by using summaries.
Logistic regression would allow the researcher to “assess how well a set of predictor variables predicts or explains categorical dependent variable” (Pallant, 2005, p. 163). The approach would assess the ‘goodness of fit’ by analyzing a number of predictor variables. Logistic regression shall show the relative significance of all predictor variables or association among variables. Pallant notes that logistic regression provides “a summary of the accuracy of the classification of cases based on the mode, which allow for the calculation of the sensitivity and specificity of the model and the positive and negative predictive values” (Pallant, 2005, p. 163).
Logistic regression shall allow the researcher to test study variables or predict categorical variables. The independent variables for the study will be categorical, continuous, or a combination of both in single analysis. The researcher will use a Forced Entry Method for analysis. In a Forced Entry Approach, the researcher shall analyze all independent variables (predictors) as a block in order to ascertain their predictive ability. While there are other methods like forward and backward, which allow the researcher to analyze a large number of independent variables, SPSS will only select variables with high predictive abilities for analysis. However, forward and backward predictive procedures are not good for analysis because differences in data can affect study outcomes.
Interpretation of the results
The researcher shall interpret results of data analysis from both descriptive statistics and logistic regression. From data features, the researcher would be able to show that there could be differences among agency nurses based on their ages and patient safety outcomes. For instance, frequencies would show the number of male and female nurses who may respond that the use of agency nurses have negative or positive patient outcomes. The researcher will also use descriptive statistics to show normality of data collected by interpreting skewness and Kurtosis test results.
SPSS would generate several results from logistic regression, but the researcher would only select outputs of interest to the study for interpretation. The researcher would use this test to accept or reject the study hypothesis. The general rule is to consider a Sig. value less than.05. This implies that the researcher would reject the null hypothesis when p is less than.05. He will accept study hypothesis to show the association between the use of agency nurses and high rates of failure to rescue, morbidity, and mortality. This result would be available in the Omnibus Tests of Model Coefficients. It will show an overall indication of ‘goodness of fit’ test. The interpretation would also consider other outputs like the chi-square value and its degree of freedom.
From logistic regression, the researcher shall focus on the type of analysis conducted, chi-square value, degree of freedom, and the p value to reflect the general model fit. He will also focus on the Model Summary with tables and percentages for correctly grouped results. There will also be pseudo R square values to show variance. The Variables in the Equation would provide a summary of study variables with 95% confidence level. The researcher shall have a verbal description of all study variables in order to help readers to understand results of the study. This would show significance of variables. Finally, logistic regression does not allow the researcher to make any assumptions based on the scores of variables and their associations. Thus, the result shall provide opportunities for further quantitative experimental studies to ascertain casual effect between variables.
Reference
Pallant, J. (2005). SPSS Survival Manual. Sydney: Ligare.