Students t-test4 p-values obtained for all analyses (for example, percent Present, Detection call accuracy, etc.) tended to distribute about 0.5, with perhaps a slight bias in favor of superior performance by the GeneChip Scanner 3000. Instances where scanner performance was different are noted in the following results.
Due to the equivalence of results between alpha and beta tests, the results described in this Technical Note are taken from the beta external tests. In addition, because conclusions reached were the same for both human probe array designs, only the HG-U133A probe array data is discussed in detail. Beta sites that deviated from the established test plan were not included in the analyses. Therefore, results from four of the six beta test sites are described.
Signal Correlation Between Scanner Types
Scatter plots of scaled Signal values for all transcripts show a high degree of similarity between scanner types and across external sites.
For this analysis, a data set from a hybridized target sample was selected as representative of all HG-U133A probe arrays. Signal values for probe sets from this data set were directly compared to the Signal values from the same probe sets at all other sites and on both scanner types. Thus, all 22,283 probe sets from the HG-U133A probe array and their Signal values were used for each correlation coefficient determination.
Correlation coefficients calculated for each scanner comparison, chip design, and site are summarized in Table 2. The high degree of similarity of Signal values between the GeneChip Scanner 3000 and GeneArray 2500 within each site is indicated by correlation coefficients approaching 1.0. The corr