Zeiss HUMPHREY 720i, Руководство пользователя

Страница: 129 / 328

STATPAC ANALYSIS & PRINTING RESULTS 7-13
REV. B 02/03 PN 51680-1 HUMPHREY FIELD ANALYZER
II -
i
series
A linear regression analysis tests the hypothesis that a slope is zero; that is, that there have
been no changes in the patient’s visual field. If this hypothesis is rejected after analysis at the p
< 5% level, the slope is said to be significant and the analysis continues at the 1% and 0.1%
levels of significance. The result is then displayed as being significant at p < 5%, p < 1%, or
p < 0.1%.
Not only the significance level but also the magnitude of the slope is important. If, for example,
the MD slope is -3.6 dB per year, plus or minus 0.9, this means that there is a 95% confidence
level that the slope is between -2.7 and -4.5 dB per year. The slope is significant at a “p” level
of less than 1%. This is a slope magnitude on the order of more than thirty times the rate of
change due to aging in the normal population. A slope of only one or two tenths of a decibel
per year would be viewed with considerably less concern, as it is similar in magnitude to the
age correction which has already been applied to the data.
If the hypothesis that the slope is zero, that is, that there has been no change in the patient’s
visual field, is not rejected, the message NOT SIGNIFICANT appears and STATPAC show a “p”
value of p > 5%. This indicates that the slope was not significant at the largest “p” value
STATPAC is programmed to consider, 5%.
The larger the number of tests analyzed, the more easily the small changes in MD are detected.
A low number of observations involves a higher risk that the analysis will fail to detect a
deterioration over time. This is the reason that STATPAC will not perform an analysis on fewer
than five test results.
The application of the linear regression analysis means that the following assumptions have
been made:
1. The true MD changes linearly with time.
2. The differences between the measured and the true MD are independent, and
identically and normally distributed.
Note: For the Change Analysis printout, you can mix tests done using the FastPac strategy
with those using Full Threshold. Because some of the STATPAC limits are slightly different
depending on whether FastPac was used or not, the significance lines for plotting PSD, SF,
and CPSD will not be on the printout if you are analyzing a series of FastPac and non-
FastPac tests. Significance lines, however, will be displayed for Mean Deviation. When SITA
tests are included, the normal box plot is not displayed for comparison (see Figure 7.6).
The Change Analysis program will also perform a regression analysis of mean deviation over
time when a series of tests using mixed strategies are used. When strategies are mixed, a
minimum of six test results are required for regression analysis; when all tests have used the
same test strategy, only five results are required. When tests are mixed with SITA results, or
SITA Standard or SITA Fast are mixed with each other, no linear regression will be performed.
Mixture of strategies is not recommended, however, as the relationship of results across
strategies is complex and difficult to quantify.
7.htp 3/14/03, 3:17 PM 13