Data - Warranty Prediction
main topics      

This command requires data to be entered in arbitrarily censored format (also called interval censored format), which provides an interval of time in which a unit failed. The Warranty Prediction command accepts arbitrarily censored data in two ways: with and without an optional frequency column.

For example, an automobile company keeps track of its shipments and warranty returns on a yearly basis. The following tables show the failure data for the past four years. It shows that 803 cars were returned during their first year of life, 790 cars were returned during their second year of life, 300 cars were returned during their third year of life, and 300 cars were returned during their fourth year of life. The missing values (*) in the End time column signify cars that had not yet failed at the end of the data collection period (in other words, cars that are still running at the present time). These rows show that 31,600 cars in their first year of life had not yet failed, 22,125 cars in their second year of life had not yet failed, 14,592 cars in their third year of life had not yet failed, and 9,490 cars had not yet failed.

Another popular format for recording warranty returns is the "triangular matrix" format. The Warranty Prediction command cannot accept this format directly, but you can convert warranty data from a triangular matrix to arbitrarily censored data with the Pre-Process Warranty Data command. In fact, the table above contains information that is identical to that in the data topic for Pre-Process Warranty Data; only the format is different.

The table above uses a frequency column to reduce the number of rows in the worksheet. However, this frequency column is optional. You can enter the same data without the optional frequency column as follows. The first column contains the row numbers.