How does Haystack work?
For each data series provided, Haystack identified the model from the model set that is the most strongly correlated. The correlation is calculated using the Pearson correlation, which is itself based on the covariance of the model and data. The results are run through several filters to ensure that certain quality control thresholds are met before results are reported.
How are the results filtered?
Haystack applied a series of different filters at different stages during the correlation calculation process. These filters are:
- Correlation Cutoff: The correlation cutoff simply excludes from output any data-to-model Pearson correlation value under the specified value.
- Fold Cutoff: The fold cutoff excludes from consideration any data series for which the fold-change between the lowest value in that series and the highest value in that series is below the specified value.
- P-value Cutoff: The program calculates p-values for each match between data and model series, based on the correlation of that match. A p-value above this cutoff results in the match not being recorded.
- Background Cutoff: Any data series which does not have two consecutive data points above this cutoff is discarded from consideration.