Example of Pre-Process Warranty Data

A reliability engineer wants to predict warranty claims that are caused by defective refrigerator compressors. The engineer collects and analyzes monthly failure data for the previous year.

Because the raw field data are recorded in a triangular matrix, the engineer uses Pre-Process Warranty Data to arrange the data into regular time-to-failure data format, which is required for the warranty analysis.

Open the sample data, CompressorFailures.MTW.
Choose Stat > Reliability/Survival > Warranty Analysis > Pre-Process Warranty Data.
From Data format, select Shipment values in a column.
In Shipment (sale) column, enter Ship.
In Return (failure) columns, enter Month1–Month12.
Click OK.

Results

Minitab reformats the original data in the worksheet into arbitrarily censored data that is grouped into intervals. The reformatted time-to-failure data is stored in 3 new columns: Start time, End time, and Frequencies, which appear directly after the original columns of data.

In this worksheet, the first 12 rows of the reformatted data (columns C14-C16) show the number of compressor returns (failures) in each monthly period following shipment. For example, the total number of returns in the first month after shipment is represented by row 1 of the worksheet, with a start time of 0 and End time of 1. A total of 7 compressors were returned in the first month after shipment.

The next 12 rows of the reformatted data show the number of compressors that were shipped but not returned in each shipment. For example, the first shipment is shown in the last row of the worksheet. Over the 12-month period since 1000 compressors were shipped in the first shipment, no compressors from that first shipment were returned. Therefore, the total number of non-returned compressors from the shipment is 1000.

C14	C15	C16
Start time	End time	Frequencies
0	1	7
1	2	5
2	3	7
3	4	10
4	5	12
5	6	4
6	7	8
7	8	6
8	9	5
9	`0	5
10	11	0
11	12	0
1	*	998
2	*	999
3	*	998
4	*	999
5	*	994
6	*	994
7	*	988
8	*	989
9	*	993
10	*	979
11	*	1000
12	*	1000

Using the reformatted time-to-failure data in the worksheet, the engineer can now evaluate the warranty claims using warranty prediction analysis.