Do you often encounter this type of data during the pharmaceutical research and development and production process?
- Dissolution curves or cell growth curves over time
- Near-infrared spectroscopy (NIR) or chromatography are commonly used in analytical chemistry.
These data are not simply "numbers," but rather "continuous curves." Traditional methods often only extract the maximum value, the area under the curve, or a specific point in time for analysis. This is not only time-consuming but also ignores crucial dynamic trends such as the curve's slope and time drift.
Try the Functional Data Explorer (FDE ) . FDE can condense complex curves into several FPC Scores ( Functional Principal Component Scores ) to capture their variations.
:pushpin: Tips: Analyze -> Specialized Modeling -> Functional Data Explorer
FDE has the following three main applications.
1. Understanding the driving factors of curve variation
Functional principal component analysis ( FDE ) identifies the main variation patterns in curves and condenses them into FPC scores . It can also be combined with design of experiments (DOE) to perform DOE analysis on the FPC scores as responses.
:pushpin: Tips: After creating the model, go to the red triangle menu next to the model -> Function Experiment Design Analysis.
2. Utilization of Functional Regression
The extracted FPC Scores are combined with regression models or machine learning. By analyzing the relationship between factors and FPC Scores , the impact on the curve can be understood.
:pushpin: Tips: After building the model, select " Save Summary" from the red triangle menu next to the Function Summary Table. This will give you the FPC Scores for each record in the original data.
3. Comparison of the curve with the target standard
By using score plots, each curve is mapped to spatial coordinates across various FPC Scores , thereby simplifying complex curve shapes to a single point on the graph. This allows for a direct comparison of shape variations between batches or differences from the target curve.
