Data Analysis

Data were analysed using statistical software, IBM SPSS version 21 and Microsoft Excel for Mac 2011 version 4.1. A 95% confidence interval was used in order to allow for a

comfortable degree of uncertainty rather than a stricter 99% confidence interval. Significance value was considered at p£0.05.

4.9.1 Descriptive Analyses

Each demographic variable was analysed using descriptive analyses, namely Frequency Distributions. Frequency distributions were important so to identify the number of times each score on a single variable occurs. The frequency and percentage distribution are outlined in a table and the results are illustrated graphically as a bar chart.

4.9.2 Factor Analysis

Factor analysis was used to investigate the relationships among the 20 variables that make up the CAMI scale (Wolff et al., 1996a). The Wolff et al. (1996a) scale suggests three constructs, namely Fear and Exclusion, Social Control and Goodwill. Factor analyses was used in order to identify different constructs or else confirm these same constructs of the original Wolff et al. (1996a) scale. Varimax rotation was used as this creates a solution in which factors are uncorrelated with each other. Abdi and Williams, (2010), describe that in Varimax rotation, the actual coordinate system is not altered, as it's the orthogonal basis that is being rotated to align with the coordinates. The same authors add that Varimax rotation maximizes the sum of the variance of the squared correlations between variables and factors (Abdi and Williams, 2010). Leech, Barrett and Morgan (2011) suggest that Varimax rotation is used to simplify the expression of a specific sub-space in terms of only a few major items. Principal component analyses produce a condensed base sub-space with many non-zero weights that make the output hard to interpret. Wolff et al. (1996a) used Varimax rotation, thus the same rotation was replicated in this study so that results can be compared to the results of the factor analyses obtained by Wolff et al. (1996a). Cronbach’s coefficient alpha test was carried out to investigate the internal consistency/reliability of the scale.

A scree plot was produced to illustrate the eigenvalues associated with the factors in descending order against the number of factors (constructs/number of factors). A scree plot

always displays a downward curve. The point where the curve levels off indicates the number of constructs (factors) that should be generated by the analysis.

The Kaiser-Meyer-Olkin (KMO) measure was used to identify if items are predicted by each factor. The KMO test is used to identify how suitable the data is for Factor Analysis, by measuring sampling adequacy for each variable in the model and also for the complete model. This statistical test measures the proportion of variance amongst variables that might be common variance. The lower the proportion the more suitable the data is for Factor Analysis. Kaiser (1974) recommends accepting values greater than 0.5. Values between 0.5 to 0.7 are considered mediocre, values between 0.7 and 0.8 are considered good, whilst those values between 0.8 and 0.9 are considered great. Any value above 0.9 are considered as superb

The Bartlett test of Sphericity was carried out to identify if the variables are correlated significantly (p<0.05) to provide a reasonable basis for the factor analyses. This test compares the observed correlation matrix to the identity matrix by checking if there is a certain redundancy between the variables that can be summarize with a few number of factors. If the variables are perfectly correlated, only one factor is sufficient. If they are orthogonal, more factors are required.

4.9.3 Analyses by variables

Each dependent variable was analysed individually against the subscales, Fear and

Exclusion, Social Control, Goodwill, and the Total Attitudinal Score. Box Plots were used

to illustrate the distribution and median of the sample population per subscale, and similarities and differences within the sample.

The box plots are followed by a bar graph representing the distribution of Fear and

Exclusion, Social Control, Goodwill and Total Attitudinal Score for each group within the

independent variable.

The Mean and Confidence interval for each group in the independent variable was calculated using SPSS and Excel and represented both in a table and as a dot plot with

error bars. This represents the mean of the groups within the independent variable, and is used to visualise any differences between the groups.

A Mann-Whiteny U test was carried out to examine differences between the independent variable Gender. The Mann-Whitney U test is a non-parametric test used to assess for significant differences in a scale or ordinal dependent variable by a single dichotomous independent variable. It is the non-parametric equivalent of the independent samples t- test. The Mann Whitney U test does not assume any properties regarding the distribution of the dependent variable in the analysis, making this test the appropriate analysis to use when analyzing dependent variables on an ordinal scale. The Mann-Whitney U-test is also the mathematical basis for the Kruskal Wallis H test. The Mann Whitney U test is also called Mann–Whitney–Wilcoxon, Wilcoxon rank-sum test, Wilcoxon–Mann–Whitney test, or Wilcoxon two-sample test. The U-test does not compare mean scores but median scores of two samples. Thus, it is much more robust against outliers and heavy tail distributions. Because the Mann-Whitney U-test is a non-parametric test, it does not require a special distribution of the dependent variable in the analysis. Therefore, it is an appropriate test to compare groups when the dependent variable is not normally distributed and at least of ordinal scale. For the test of significance of the Mann-Whitney U-test, it is assumed that with a large sample size, the distribution of the U-value approximates a normal distribution. The U-value calculated with the sample can be compared against the normal distribution to calculate the confidence level. The U-value represents the number of times observations in one sample precede observations in the other sample in ranking.

Conversely the Kruskal-Wallis H test was conducted in order to examine the differences between variables with three or more groups, namely, Grade, Age, Nursing Education, Years in Nursing Service, Work Setting, Years in Mental Health Setting, Mental Health Work Setting and Contact with Mental Health Service Users. The Kruskal-Wallis H test is

a nonparametric method used for testing whether samples originate from the same

distribution. It is used for comparing two or more independent samples of equal or different sample sizes. It extends the Mann-Whitney U test when there are more than two groups. The parametric equivalent of the Kruskal-Wallis test is the one-way ANOVA. Both tests assess for significant differences on a continuous dependent variable by a categorical independent variable (with two or more groups). In the ANOVA, it is assumed that the dependent variable is normally distributed and there is approximately equal

variance on the scores across groups. However, when using the Kruskal-Wallis Test, such assumptions are not required. Therefore, the Kruskal-Wallis test can be used for both

continuous and ordinal-level dependent variables. The Kruskal-Wallis test does not

identify where this stochastic dominance occurs or for how many pairs of groups stochastic dominance obtains. Therefore, in order to analyse the specific sample pairs for stochastic dominance in post hoc testing the Mann Whitney U test was used. Results for Post Hoc analyses are presented as p values.

For the independent variables Age, Years in Nursing Service and Years in Mental Health

Work Setting, the Spearman rank correlation test was used to assess the relationship

between the variable, such as Age, and each subscale, Fear and Exclusion, Social Control,

Goodwill and Total Attitudinal Score. Spearman’s correlation coefficient is a statistical test

that measures the strength and direction of association between two ranked variables. Similar to the Pearson correlation test, the Spearman test has a value between +1 and −1. The closer the correlation coefficient (rs) value is to +1/-1 the stronger the relationship.

The direction of the association is denoted by the positive (+) or a negative (-) sign of the

rs value. If rs is positive, it means that as one variable gets larger the other gets larger. If r

is negative it means that as one gets larger, the other gets smaller also known as an "inverse" correlation. This test is a non-parametric test as it does not require a normal distribution, however the Pearson test cannot be used in this analyses as data for variables Age, Years in Nursing Service, Years in Mental Health Setting and Contact with Mental Health Service Users is ordinal. This justifies the use of the Spearman rank correlation test as opposed to the parametric Pearson correlation test.

To conclude the data analyses, a regression was carried out to identify significant predictors contributing to the primary dependent variable, that is, Total Attitude Score.