### Preview

This step performs a simple linear regression for each batch to verify the model validity and the data consistency (capture error,…).

- Click on
**Validate**to launch calculation for the parameter.

**Note**: It is possible to modify data and parameter information at this step. Parameter data can be modified directly on the grid. Parameter information can be modified via the **Edit** button.

#### The kinetic is not linear

If the linear kinetic assumption is not fulfilled, it is possible to transform data by applying a log10 transformation.

- Select Logarithm (resp. Linear) to apply log10 transformation (resp. to cancel log10 transformation)

**Note**: The transformation is applied only if all data of the studied parameter are strictly greater than zero.

**Tip**: The transformation is visible on the Y axis of the preview graph.

#### What about “<LOQ” values

Values noted “<LOQ” can be replaced by a numeric value. The replacement value is then used in the statistical analysis.

**Note**: If a negative or 0 value is entered on a log transformed parameter, the parameter is automatically switched to the linear default.

**Tip**: Values noted “<LOQ” can be replaced by LOQ/2 or LOQ/sqrt(2). The replacement is visible on the data grid.

### Shelf-life computation

The ICH recommendations are applied for the shelf-life calculation.

In presence of more than one batch, the poolability of data is tested with a significance of 25%.

According the final scenario from the above algorithm, the final shelf-life is calculated according to the ICH recommendations. For reminder, a 95% lower (resp. upper) confidence bound is used for a parameter known to decrease (resp. increase) in time. For a parameter that can either increase or decrease with time, a 95% two-sided confidence interval is used.

**Note**: If the data from all batches cannot be combined (SSSI or CSCI), the final shelf-life is the minimum of shelf-lives.

**Note**: Shelf-lives are floor rounded and are report as:

- “>15” with year unit when the estimated shelf-lives are greater than 15 years
- “>150” with month unit when the estimated shelf-lives are greater than 150 months
- “>100” with week unit when the estimated shelf-lives are greater than 100 weeks
- “>350” with day unit when the estimated shelf-lives are greater than 350 days
- “>48” with day unit when the estimated shelf-lives are greater than 48 hours

#### Quality of the fit

Residuals are plotted against time point. You typically want residuals randomly scattered around zero.

#### Release limit

Release limits are computed automatically. They ensure that the claimed shelf-life is achieved.

**Tip**: The estimated shelf-life is calculated whereas the claimed shelf-life is the desired one.