SPC/Statistics > Data Tests

MeasurLink can check data as it is received for statistical patterns that indicate a change in randomness. A change in randomness can indicate that the manufacturing process has changed from its initial settings. The initial settings are designed to create the best Part possible, so when the system deviates from the initial settings the manufacturing process has moved off its optimum settings.

Statistical Data Tests

The statistical tests are often referred to as Shewart Tests or Western Electric Tests.
 

Extreme Point (Out of Control)

Subgroup average falls outside the control limits. This may indicate a shift in the process.Extreme Point
For normally distributed random data this occurs rarely at 0.27% of the time.
 

Run Above or Below Centerline

Out of control condition - specified number of consecutive subgroup averages (typically 7, 8, 9) are all above or below the center line (overall average of the data). This may indicate a shift in the process.Above or Below Centerline
For data with a symmetrical distribution about the X-DBar value:
  • 7 data points occur 1.56% of the time
  • 8 data points occur 0.78% of the time
  • 9 data points occur 0.39% of the time

Linear Trend

Out of control condition - specified number of consecutive subgroup averages (typically 6) are all increasing or decreasing in value. Neither the chart centerline nor the zones are considered. This could be a sign of tool wear.
Linear Trend
For normally distributed random data with 7 data points, all increasing or all decreasing occurs 0.28% of the time.
 

Oscillatory Trend

Out of control condition is when a specified number of consecutive subgroup averages (typically 14) alternate up and down. Neither the chart centerline nor the zones are considered. This may indicate trending toward out of control process.
Oscillatory Trend
For data with a normal probability distribution this occurs 0.46% of the time.
 

2 of 3 Beyond 2 Sigma

Out of control condition - 2 of 3 consecutive subgroup averages are in Zone A or beyond (2 standard deviations or beyond) on the same side of Center line (overall average) and the 3rd subgroup average is one of the faulting subgroups. This may indicate a shift in the process.
2 of 3 Beyond Sigma
For normally distributed random data this occurs 0.15% of the time.
 

4 or 5 Beyond 1 Sigma

Out of control condition - 4 of 5 consecutive subgroup averages are in Zone B or beyond (1 standard deviation or beyond) on the same side of Center line (overall average) and the 5th subgroup average is one of the four faulting subgroups. This may indicate a shift in the process.
2 of 3 Beyond Sigma
For normally distributed random data this occurs 0.28% of the time.
 

Stratification

A specified number of consecutive subgroup averages (typically 15) are all within Zone C (within 1 standard deviation of overall average). This could indicate a reduction of variation in the process.
Stratification
For normally distributed random data 15 data points’ display oscillatory behavior 0.46% of the time.
 

Mixing / Over Control

Out of control condition - specified number of subgroup averages (typically 8) are outside Zone C (outside 1 standard deviation of average) on both sides of the Center line (overall average). This may indicate multiple processes (e.g. multiple machines) being plotted on a single chart or hyper-adjustment of the process.
Mixing/Overcontrol
For normally distributed random data 8 data points’ display oscillatory behavior 0.46% of the time.

Non-Statistical Data Tests

Out of Tolerance

This is not a control test, but a test of observations meeting specification.
Out of Tolerance

2 of 3 in Yellow (Precontrol Chart)

Precontrol is not a control test, but based on specification. Green area is near target (nominal), Yellow outside Green area but still within specification, and Red outside of tolerance. Test failure is when a single observation falls within Red area or 2 of 3 consecutive observations are in Yellow and the last of the 3 is in Yellow. This test is typically used to alert Operator to re-center process.
Precontrol Test