Desarrollo de las actividades.

From the logic structure it is derived that there are 3 parts in the analysis:

1. Analysis to examine the effect of occupants perception of the appearance from the work position on the occurrence of eye symptoms.

Analysis to examine the effect of the measured physical light attributes on the occurrence of eye symptoms.

Analysis to examine the effect of measured physical light attributes on occupants perception of the appearance from the work position.

2.

3.

The statistical computation used to examine data was carried out using "Stat View" programme.

To carry out the analysis, it is necessary to first analyse the affect of work position orientation on;

i. occurrence of eye symptoms,

iii. measured physical light attributes.

Each of the above involve two separate analyses^^; • mean and standard deviation comparison, and • Kruskal-Wallis non-parametric analysis.

5 . 7 . 1

Comparison Of The Mean Scores

To checked the variation in the variable, a comparison of the mean scores of the variable was carried out. This was done by simply comparing the mean of the variable assessed for the groups, and rank them.

5 . 7 . 2

Kruskal-Wallis Non-Parametric Analysis

The Kruskal-Wallis analysis was carried out to see if the differences observed in the comparison of the mean scores is significant. The Kruskal-Wallis non-parametric analysis was used because the data was suspected to be askewed, has a small sample size, and the number of group is more than 2.

The test involves a comparison of the rankings for each of the categories of the nominal scale variable. The 18 samples were divided according to the seat position with respect to the window; facing the window, with the window to the side or with the window to the back.

The procedure for the analysis is as follows: first the samples are divided according to the orientation. There are three groups: facing the window, with the window to the side or with the window to the back. The orientation is assigned an x variable, and the appearance is assigned a y variable. From the non-parametric analysis, the distribution value "H" is computed. This value determines whether the variables under study vary significantly between different orientations. To determine the significance of H, the chi-square test is used because the sampling distribution of H approximates chi-square distribution. The aim of the chi-square test is to test for significant differences. To determine H, the significance level and the degree of freedom is required. The degree of freedom depends on the number of groups. The degree of freedom is computed using k-1, where k is the number of groups compared (The degree of freedom gives the

lighting studies advance statistical methods for data analysis are rarely required - the analyses to some extent require intutition and the results are open to interpretation. Architectural experiments rarely require a precision greater than that given by evaluation o f median for an expression o f the average and by the interquartile range for the expression o f the distribution about the mean (Hopkinson, 1963).

row to refer to in the chi-square table). The desired level of significance, p, which is taken to be appropriate here as 0.05. The value of H is then compared with the critical values from the chi-square table. In order to reject the null hypothesis, H must be equal or larger than the critical value - Hq (Sarantakos, 1994; Blalok, 1979). But in "Stats View", as indicated in its manual, variables that vary significantly according to the sub­ division are automatically given by the significance level with corrections for ties^^. Therefore, variables that have p < 0.05 vary significantly with orientation. In the table, the factors that vary significantly are indicated by bold prints.

Conditions for Chi-square test; 1. Independent random samples.

2. It is required to have at least 5 samples per group (cell). Should a group be any smaller (thin cell), this ought to represent only 25%. In other words, if there are four groups and one group has less than 5 samples, the percentage this group constitutes of the total group is 20%. As this does not exceed the 25% it is possible to carry out the chi-square test.

5 . 7 . 3

Significance Level

In social science studies, the significance levels of 0.05, 0.02 and 0.01 are recommended. The significance level gives an idea of the accuracy of the study. The smaller the significance level the more confidence in the observed association and reliability of the finding. Of the three levels stipulated, 0.05 is commonly used and normally accepted. A significance level of 0.05 means that only 5 out of a hundred samples would come out by chance with the association observed in the sample. It is the general consensus that this level gives credibility and this study has adopted this figure as its level of significance. This level is often used in experimental studies. In natural settings it is possible that this level should be raised to 0.1 (the significance level is also dependent on the sample size used in the study, for example if only 10 sample is used in the study a significance level of 0.01 is more appropriate whereas if the sample size is

1000 a significance level of 0.1 is permissible.