Assessing Capability

Joel Smith Commercial Sales Minitab, Inc.

Schedule
Learn about the tools Two “continuous” examples
• Assessment of cookout locations • In class example

One “binomial” example One “count” example (if time permits)

Brain Warmer
The Monty Hall Show

Brain Warmer Pick 1 2 3 Shown Should 2/3 3 2 Stay Move Move Pick 1 2 3 Shown Should 3 1/3 2 Move Stay Move Pick 1 2 3 Shown Should 2 1 1/2 Move Move Stay ⅔ ⅔ ⅔ .

What is “Capability”? Assess quality Quantify ability to meet specifications Distinguish short.and long-term .

Data Types Continuous • Length • Time • Temperature Binary • Yes/No. Pass/Fail • How many heads in X coin flips Count • Defects/part • Orders in a day .

Boxplot Determine distribution of data • Probability Plot Evaluate stability • Control Charts Capability • Capability Analysis . Attribute Gage R&R Collect data Look at the data • Histogram.Assessing Capability Determine specifications Verify Measurement System • Gage R&R.

Assessing Capability Determine specifications • Avoid this topic here… Verify Measurement System • Error types • Ability to measure accurately Collect data • Short-term and long-term • Subgrouping • Randomize collection of data .

Assessing Capability Look at the data • Always! Determine distribution of data • Most data is not normal • Good fit is critical • Some data have natural distribution Evaluate stability • Unstable process is unpredictive Use distribution to quantify capability • Capability quantified using several statistics .

Two Examples I want to plan a 4th of July cookout • Where should I have it? • What factors should I consider? • How likely is each location to satisfy my requirements? We make high-strength cord used to secure parachutes • How long is each cord? • How likely is each cord to be within my specs? .

Two Examples 4th of July Cookout is in slides • My locations: – State College. PA – Pasadena. CA • My factors – Temperature – Precipitation Cord will be done here • Evaluate Length .

. PA What is the capability of State College to produce good weather on July 4th? Average Temperature should be between 65 and 85 Precipitation should be <0.1 in.Cookout: State College.

verify measurement system • Collect data • Look at the data • Evaluate stability using a Control Chart • Determine the distribution • Perform a Capability Analysis . PA To assess capability: • If necessary.Cookout: State College.

etc.) • Attribute (Yes/No. Temperature. Time. Poor/Fair/Good.) Establishes how much variability is coming from parts versus operators .Measurement System Analysis (MSA) Do prior to collecting and analyzing data Two types: • Continuous (Length. etc.

All others bring data .Measurement System Analysis (MSA) In God we trust.

Cord: MSA in Class To test whether our Measurement System is sufficient: • 3 Operators (volunteers?) • 6 Parts • 2 Measurements per part per operator Randomize! We will do Attribute Gage R&R later .

Cord: MSA in Class (Do MSA now…) .

Cookout: State College. no MSA will be done Data has already been collected Next steps: • • • • Look at the data Evaluate stability using a Control Chart Determine the distribution Perform a Capability Analysis . PA For our weather data.

089 .Cookout: State College.000 74.000 9 5 % C onfidence Inter vals Mean Median 68.965 55 60 65 70 75 80 85 95% C onfidence Interv al for M ean 95% C onfidence Interv al for M edian 95% C onfidence Interv al for S tDev 5.888 5.881 34.582 0.221 6.0008702 80 55.0962166 -0.0 0.0 70.5 70.870 69.000 66.5 71.000 70.20 0.0 69.000 85.196 71. PA Look at the data: Summary for TAVE (F) A nderson-D arling N ormality Test A -S quared P -V alue M ean S tD ev V ariance S kew ness Kurtosis N M inimum 1st Q uartile M edian 3rd Q uartile M aximum 68.0 68.579 68.000 71.5 69.

888 5.000 70.579 68.0 68.000 74.582 0.5 69.5 70.0008702 80 55.000 85. PA Data appears “normal” Symmetry 55 60 65 70 75 80 85 Summary for TAVE (F) A nderson-D arling N ormality Test A -S quared P -V alue M ean S tD ev V ariance S kew ness Kurtosis N M inimum 1st Q uartile M edian 3rd Q uartile M aximum 68.196 71.0 70.965 95% C onfidence Interv al for M ean Mean ~ Median 95% C onfidence Interv al for M edian 95% C onfidence Interv al for S tDev 5.089 .881 34.0 69.20 0.870 69.000 71.0 0.5 71.000 9 5 % C onfidence Inter vals Mean Median 68.0962166 -0.000 66.221 6.Cookout: State College.

89 U C L=23. PA Evaluate Stability using a Control Chart: I-MR Chart of TAVE (F) 90 U C L=88.Cookout: State College.22 20 M oving Range 15 10 5 0 1 9 17 25 33 41 O bser vation 49 57 65 73 LC L=0 __ M R=7.11 .79 Individual V alue 80 70 60 50 1 9 17 25 33 41 O bser vation 49 57 65 73 LC L=50.99 _ X=69.

99 _ X=69.79 80 70 60 50 1 9 17 25 33 41 O bser vation 49 57 65 73 LC L=50.89 U C L=23.11 .22 20 M oving Range 15 10 5 0 1 9 17 25 33 41 O bser vation 49 57 65 73 LC L=0 __ M R=7. PA Evaluate Stability using a Control Chart: I-MR Chart of TAVE (F) 90 Individual V alue Random.Cookout: State College. stable No “out of control” Location/Spread U C L=88.

PA Determine Distribution What is the “Normal” distribution? Other distributions: • Weibull • Largest/smallest extreme value • Exponential .Cookout: State College.

881 80 0. PA Use a Probability Plot to Determine Distribution: Probability Plot of TAVE (F) Normal 99.9 99 95 90 80 70 60 50 40 30 20 10 5 1 0.1 Mean StDev N AD P-Value 69.89 5.204 0.870 Percent 50 60 70 TAVE (F) 80 90 .Cookout: State College.

PA Use a Probability Plot to Determine Distribution Probability Plot of TAVE (F) “Fat Pencil” test “Squinty Eye” test Anderson-Darling • P-value Normal 99.Cookout: State College.881 80 0.204 0.870 Percent 50 60 70 TAVE (F) 80 90 .1 Mean StDev N AD P-Value 69.89 5.9 99 95 90 80 70 60 50 40 30 20 10 5 1 0.

perform Capability Analysis • Specs: 65 to 85 degrees Key assumptions: • Data is from a stable process • Data is well-fit by distribution We will learn: • Characteristics of data • Likelihood of “bad” parts • Short-term vs.Cookout: State College. PA Finally. Long-term performance .

28 0.8875 S ample N 80 S tD ev (Within) 6. PA Finally.23 .57 0.Cookout: State College.82 C pk 0.23 Total 220877.53 > USL 7083.36 P P M Total 208904.00 PPM > USL 0.75 E xp.26 O v erall C apability Pp PPL PPU P pk C pm 0.85 0.28 * 55 O bserv ed P erformance P P M < LS L 187500.8993 USL Within Ov erall P otential (Within) C apability Cp 0.00 E xp.16079 S tD ev (O v erall) 5.54 C P L 0.86 PPM > USL 5207.00 P P M Total 187500. perform Capability Analysis Process Capability of TAVE (F) LSL P rocess D ata LS L 65 Target * USL 85 S ample M ean 69. O v erall P erformance P P M < LS L 203696.26 C P U 0. PPM PPM PPM 60 65 70 75 80 85 Within P erformance < LS L 213794.

26 C P U 0.8875 S ample N 80 S tD ev (Within) 6.23 Total 220877.28 * Characteristics of Data55 60 O bserv ed P erformance P P M < LS L 187500.00 E xp.23 .86 PPM > USL 5207.00 PPM > USL 0.Cookout: State College.8993 USL Within Ov erall P otential (Within) C apability Cp 0.53 > USL 7083.36 P P M Total 208904. perform Capability Analysis Process Capability of TAVE (F) LSL P rocess D ata LS L 65 Target * USL 85 S ample M ean 69.26 O v erall C apability Pp PPL PPU P pk C pm 0.54 C P L 0.75 E xp.00 P P M Total 187500.82 C pk 0. PA Finally.28 0.57 0.85 0. O v erall P erformance P P M < LS L 203696.16079 S tD ev (O v erall) 5. PPM PPM PPM 65 70 75 80 85 Within P erformance < LS L 213794.

26 C P U 0.85 0.8875 S ample N 80 S tD ev (Within) 6.8993 USL Within Ov erall P otential (Within) C apability Cp 0.23 Total 220877.82 C pk 0.53 > USL 7083.54 C P L 0. O v erall P erformance P P M < LS L 203696. PPM PPM PPM 60 65 70 75 80 85 Within P erformance < LS L 213794.86 PPM > USL 5207.00 P P M Total 187500.75 E xp.Cookout: State College.28 * Likelihood of “bad” parts 55 O bserv ed P erformance P P M < LS L 187500.16079 S tD ev (O v erall) 5.00 PPM > USL 0.36 P P M Total 208904. PA Finally.28 0.57 0.23 .00 E xp. perform Capability Analysis Process Capability of TAVE (F) LSL P rocess D ata LS L 65 Target * USL 85 S ample M ean 69.26 O v erall C apability Pp PPL PPU P pk C pm 0.

00 P P M Total 187500. O v erall P erformance P P M < LS L 203696.00 E xp.16079 S tD ev (O v erall) 5. PA Finally.54 C P L 0.82 C pk 0.36 P P M Total 208904.75 Short-term vs.26 C P U 0. 70 75 80 85 Long-term E xp.28 * 55 O bserv ed P erformance P P M < LS L 187500.86 PPM > USL 5207.85 0.8993 USL Within Ov erall P otential (Within) C apability Cp 0.23 .57 0.23 Total 220877.53 > USL 7083.26 O v erall C apability Pp PPL PPU P pk C pm 0.Cookout: State College.28 0.8875 S ample N 80 S tD ev (Within) 6. perform Capability Analysis Process Capability of TAVE (F) LSL P rocess D ata LS L 65 Target * USL 85 S ample M ean 69. PPM PPM PPM 60 65 Within P erformance < LS L 213794.00 PPM > USL 0.

We will follow the same procedure: • • • • Look at the data Evaluate stability Determine distribution Perform Capability Analysis .Cookout: State College.1 in. PA Now let’s take a look at Precipitation Recall we want <0.

00000 0.21408 0.00000 0.6 0.2 0.05593 2.18528 .06261 0.4 0.Cookout: State College.95297 80 0. PA Look at the data: Summary for PRCP (in) A nderson-D arling N ormality Test A -S quared P -V alue < M ean S tD ev V ariance S kew ness Kurtosis N M inimum 1st Q uartile M edian 3rd Q uartile M aximum 0.07750 0.04583 2.125 0.150 15.075 0.050 0.00 0.005 0.8 95% C onfidence Interv al for M ean 95% C onfidence Interv al for M edian 95% C onfidence Interv al for S tDev 0.25357 -0.11025 0.00000 0.025 0.00000 9 5 % C onfidence Inter vals Mean Median 0.100 0.15789 0.0 0.82000 0.000 0.02000 0.

00000 9 5 % C onfidence Inter vals Mean Median 0.25357 95% C onfidence Interv al for M ean Mean ≠ Median 95% C onfidence Interv al for M edian 95% C onfidence Interv al for S tDev 0.2 0.18528 .11025 0.100 0.07750 0.125 0.025 0.02000 0.4 0.82000 0.00 0.00000 0.005 0.Cookout: State College.21408 0.150 15.05593 2.8 Summary for PRCP (in) A nderson-D arling N ormality Test A -S quared P -V alue < M ean S tD ev V ariance S kew ness Kurtosis N M inimum 1st Q uartile M edian 3rd Q uartile M aximum 0.0 0.000 0.075 0.050 0.15789 0.06261 0.00000 0.04583 2.95297 80 0. PA Data appears “skewed” No symmetry -0.00000 0.6 0.

Cookout: State College.6 0. PA Evaluate Stability using a Control Chart: I-MR Chart of PRCP (in) 0.8 1 1 1 1 1 1 1 1 1 1 Individual V alue 0.4 11 11 0.2 0.0 1 9 17 25 33 41 O bser vation 49 57 65 73 U C L=0.2675 _ X=0.8 11 1 1 11 11 1 11 11 11 1 1 11 M oving Range 0.0591 LC L=0 .1932 __ M R=0.6 0.4 1 1 1 0.1103 LB=0 0.2 0.0 1 9 17 25 33 41 O bser vation 49 57 65 73 U C L=0.

4 1 1 1 0.6 0.6 0.2675 _ X=0.1103 LB=0 0.0 1 9 17 25 33 41 O bser vation 49 57 65 73 U C L=0.2 0.8 M oving Range 0.2 0.0 1 9 17 25 33 41 O bser vation 49 57 65 73 U C L=0.Cookout: State College.0591 LC L=0 . PA Evaluate Stability using a Control Chart: I-MR Chart of PRCP (in) 0.8 Individual V alue 1 1 1 1 1 1 1 1 1 1 Unstable “Out of control” 0.4 11 11 1 1 11 11 1 11 11 11 1 1 11 Non-predictive 11 0.1932 __ M R=0.

Cookout: State College. PA STOP! .

PA Assumptions are not met No stability = No capability At this point: • Special causes • Other factors .Cookout: State College.

Cookout: State College. PA Is State College a good location? Temperature • Average temperature stable year-to-year • Normal distribution • 79% chance of “good” Precipitation • Precipitation is unstable • Cannot determine capability .

Cookout: Pasadena. CA How about Pasadena? Evaluate same criteria • Temperature (65 to 85) Precipitation (<0.) Remember our process: • • • • • Verify measurement system. Collect data (MSA) Look at the data (Histogram) Evaluate stability (Control Chart) Determine distribution (Probability Plot) Perform Capability Analysis .1 in.

000 74.0 1.5 73.949 .78 0.000 76.000 9 5 % C onfidence Inter vals Mean Median 71.043 73.0 72.000 86. CA Look at the data: Summary for TAVE (F) A nderson-D arling N ormality Test A -S quared P -V alue < M ean S tDev V ariance S kew ness Kurtosis N M inimum 1st Q uartile M edian 3rd Q uartile M aximum 71.896 0.000 70.857208 0.0 73.000 72.Cookout: Pasadena.571 20.427 64 68 72 76 80 84 95% C onfidence Interv al for M ean 95% C onfidence Interv al for M edian 95% C onfidence Interv al for S tDev 3.462320 78 64.013 4.5 74.000 5.005 73.982 71.5 72.0 71.

000 74.000 5.0 72.005 73.5 72.5 74.000 9 5 % C onfidence Inter vals Mean Median 71.000 72.000 86.571 20.0 73.78 0.000 70.949 .000 76.0 1.0 71.013 4.462320 78 64.043 73.5 73.857208 0.982 71.896 0.427 Mean ≠ Median 95% C onfidence Interv al for M ean 95% C onfidence Interv al for M edian 95% C onfidence Interv al for S tDev 3.Cookout: Pasadena. CA Data appears partly skewed Some Asymmetry 64 68 72 76 80 84 Summary for TAVE (F) A nderson-D arling N ormality Test A -S quared P -V alue < M ean S tDev V ariance S kew ness Kurtosis N M inimum 1st Q uartile M edian 3rd Q uartile M aximum 71.

Cord: Look at data First we need to collect data Need a good distribution fit • Generally 25-50 points Use Histogram .

99 _ X=73.00 __ M R=4.Cookout: Pasadena.03 Individual V alue 78 72 66 60 1 9 17 25 33 41 O bser vation 49 57 65 73 LC L=59.01 20 15 10 5 0 1 9 17 25 1 M oving Range U C L=16. CA Evaluate Stability using a Control Chart: I-MR Chart of TAVE (F) 84 U C L=86.90 LC L=0 33 41 O bser vation 49 57 65 73 .

01 20 15 10 5 0 1 9 17 25 1 M oving Range U C L=16.00 __ M R=4.03 Mostly stable One “out of control” Can we proceed? 78 72 66 60 1 9 17 25 33 41 O bser vation 49 57 65 73 LC L=59. CA Evaluate Stability using a Control Chart: I-MR Chart of TAVE (F) 84 Individual V alue U C L=86.90 LC L=0 33 41 O bser vation 49 57 65 73 .Cookout: Pasadena.99 _ X=73.

Cord: Evaluate Stability (Evaluate stability using Control Chart…) .

Out of Control Points Many out of control points: • Unstable process • Special causes • Other factors Very few out of control: • Look for special cause • Only if legitimate. remove .

782 <0.571 78 1.005 Percent 60 65 70 75 TAVE (F) 80 85 90 .Cookout: Pasadena. CA Use a Probability Plot to Determine Distribution: Probability Plot of TAVE (F) Normal 99.01 4.1 Mean StDev N AD P-Value 73.9 99 95 90 80 70 60 50 40 30 20 10 5 1 0.

9 90 50 Weibull .9 99 99.9 99 3-P arameter Loglogistic .782 P -V alue < 0.491 P -V alue < 0.1 40 60 T A V E ( F) 80 Largest E xtreme V alue .9 99 90 99.1 10 1 0.95% C I 99.010 Largest E xtreme V alue A D = 0.1 1 10 T A V E ( F) .95% C I 99.005 Weibull A D = 3.544 P -V alue = 0. CA Use Probability Plots to Determine Distribution: Probability Plot for TAVE (F) N ormal .174 3-P arameter Loglogistic A D = 0.95% C I G oodness of F it Test N ormal A D = 1.T hr eshold 100 90 50 10 0.556 P -V alue = * P er cent P er cent 60 70 80 T A V E ( F) 90 50 10 1 0.95% C I P er cent P er cent 70 80 90 T A V E ( F) 100 90 50 10 1 0.1 .Cookout: Pasadena.

CA Use a Probability Plot to Determine Distribution Probability Plot of TAVE (F) Largest Extreme Value .1 Loc Scale N AD P-Value 70.544 0.174 Percent 60 70 80 TAVE (F) 90 100 .95% CI “Fat Pencil” test “Squinty Eye” test Anderson-Darling • P-value 99.94 3.9 99 98 97 95 90 80 70 60 50 40 30 20 10 1 0.Cookout: Pasadena.628 78 0.

Cord: Determine Distribution (Determine distribution using Probability Plot…) .

perform Capability Analysis • Specs: 65 to 85 degrees Key assumptions:  Data is from a stable process  Data is well-fit by distribution We will learn: • Characteristics of data • Likelihood of “bad” parts • Long-term performance only . CA Finally.Cookout: Pasadena.

03 Process Capability of TAVE (F) USL O v erall C apability Pp 0.Cookout: Pasadena.30 64 68 72 76 80 84 88 .65 PPL 0. O v erall P erformance P P M < LS L 5877.89 PPU 0.9361 S cale 3.56 P pk 0.51 P P M Total 25641.56 E xp. perform Capability Analysis Calculations Based on Largest Extreme Value Distribution Model LSL P rocess D ata LS L 65 Target * USL 85 S ample M ean 73.92 P P M > U S L 20500.38 P P M Total 26378.62756 O bserv ed P erformance P P M < LS L 12820.0128 S ample N 78 Location 70. CA Finally.51 P P M > U S L 12820.

56 E xp.30 Characteristics of Data 64 68 72 76 80 84 88 .9361 S cale 3.56 P pk 0. O v erall P erformance P P M < LS L 5877. PA Finally.89 PPU 0.62756 O bserv ed P erformance P P M < LS L 12820. perform Capability Analysis Calculations Based on Largest Extreme Value Distribution Model LSL P rocess D ata LS L 65 Target * USL 85 S ample M ean 73.0128 S ample N 78 Location 70.38 P P M Total 26378.51 P P M Total 25641.Cookout: State College.51 P P M > U S L 12820.03 Process Capability of TAVE (F) USL O v erall C apability Pp 0.65 PPL 0.92 P P M > U S L 20500.

51 P P M Total 25641. PA Finally.62756 O bserv ed P erformance P P M < LS L 12820.56 E xp.65 PPL 0.89 PPU 0.30 Likelihood of “bad” parts 64 68 72 76 80 84 88 .0128 S ample N 78 Location 70.03 Process Capability of TAVE (F) USL O v erall C apability Pp 0.9361 S cale 3.38 P P M Total 26378.51 P P M > U S L 12820.56 P pk 0. O v erall P erformance P P M < LS L 5877. perform Capability Analysis Calculations Based on Largest Extreme Value Distribution Model LSL P rocess D ata LS L 65 Target * USL 85 S ample M ean 73.92 P P M > U S L 20500.Cookout: State College.

9361 S cale 3.03 Process Capability of TAVE (F) USL O v erall C apability Pp 0.56 E xp.89 PPU 0.51 P P M Total 25641.51 P P M > U S L 12820. perform Capability Analysis Calculations Based on Largest Extreme Value Distribution Model LSL P rocess D ata LS L 65 Target * USL 85 S ample M ean 73.65 PPL 0. PA Finally.92 P P M > U S L 20500.Cookout: State College.30 Long-term only 64 68 72 76 80 84 88 .62756 O bserv ed P erformance P P M < LS L 12820.0128 S ample N 78 Location 70.56 P pk 0. O v erall P erformance P P M < LS L 5877.38 P P M Total 26378.

Cord: Capability Analysis (Perform Capability Analysis now…) .

1 in. Collect data (MSA) Look at the data (Histogram) Evaluate stability (Control Chart) Determine distribution (Probability Plot) Perform Capability Analysis . CA How about Precipitation? Evaluate same criteria • Precipitation (<0.Cookout: Pasadena.) Remember our process: • • • • • Verify measurement system.

06 95% C onfidence Interv al for M ean 95% C onfidence Interv al for M edian 95% C onfidence Interv al for S tDev 0.Cookout: Pasadena.005 0.0000 0.8318 78.000000 0.05 0.002301 0.000046 8.000000 0.060000 0.0010 0.00 0.03 0.000000 0.000769 0.0020 29.01 0.000000 9 5 % C onfidence Inter vals Mean Median -0.0005 0.0005 0. CA Look at the data: Summary for PRCP (in) A nderson-D arling N ormality Test A -S quared P -V alue < M ean S tDev V ariance S kew ness Kurtosis N M inimum 1st Q uartile M edian 3rd Q uartile M aximum -0.000763 0.0015 0.000000 0.008066 0.000000 0.04 0.005869 .19 0.006794 0.0010 -0.02 0.0000 78 0.

0020 Summary for PRCP (in) A nderson-D arling N ormality Test A -S quared P -V alue < M ean S tDev V ariance S kew ness Kurtosis N M inimum 1st Q uartile M edian 3rd Q uartile M aximum -0.02 0.000769 0.000000 0.19 0.Cookout: Pasadena.000046 8.0000 0.008066 0.002301 0. CA Unusual data Nearly all values equal or nearly equal Mean ≠ Median Mean Median -0.03 0.0010 0.000763 0.0015 0.00 0.005 0.000000 9 5 % C onfidence Inter vals 0.005869 29.06 95% C onfidence Interv al for M ean 95% C onfidence Interv al for M edian 95% C onfidence Interv al for S tDev .060000 0.000000 0.006794 0.8318 78.0005 0.05 0.04 0.000000 0.0010 -0.0000 78 0.01 0.0005 0.000000 0.000000 0.

00509 UC M R=0.030 0.000 1 9 17 25 33 41 O bser vation 49 57 65 73 _ C L=0.045 0.00156 LC L=0 .000 1 9 17 25 33 41 O bser vation 49 57 65 73 __ L=0.00338 0.00491 U X=0.00077 LC L=-0. CA Evaluate Stability using a Control Chart: I-MR Chart of PRCP (in) 0.060 1 Individual V alue 0.Cookout: Pasadena.045 0.015 0.060 11 M oving Range 0.030 0.015 0.

060 Individual V alue 1 Mostly stable? One “out of control” 0.00491 U X=0.060 M oving Range 0.00509 UC M R=0.000 1 9 17 25 33 41 O bser vation 49 57 65 73 _ C L=0.Cookout: Pasadena.015 0.000 1 9 17 25 33 41 O bser vation 49 57 65 11 Can we proceed? __ L=0.045 0.045 0.030 0.015 0.00338 0. CA Evaluate Stability using a Control Chart: I-MR Chart of PRCP (in) 0.030 0.00156 LC L=0 73 .00077 LC L=-0.

Out of Control Points Many out of control points: • Unstable process • Special causes • Other factors Very few out of control: • Look for special cause • Only if legitimate. remove .

CA Our process is out of control: Binary process? Can evaluate capability for binary (later) Not enough data… .Cookout: Pasadena.

CA STOP! .Cookout: Pasadena.

Cookout: Pasadena, CA
Is Pasadena a good location? Temperature
• Average temperature stable year-to-year • Non-normal distribution • 97.4% chance of “good”

Precipitation
• Precipitation is unstable • Cannot determine capability

Cookout Comparison
Which city is better Temperature
• State College = 79% • Pasadena = 97.4%

Precipitation
• Precipitation is unstable for both • Relative rate is much lower in Pasadena

Cookout Comparison
A quick graph:
Boxplot of TMIN (F), TAVE (F), TMAX (F)
110 100 90 80 70 60 50 40
City Pasadena State College TMIN (F) Pasadena State College TAVE (F) Pasadena State College TMAX (F)

Data

Pill Quality Example We work for a pharma company. and pill quality is critical Need to evaluate our capability Remember: • • • • • • Verify measurement system Collect data Look at the data Evaluate stability using a Control Chart Determine the distribution Perform a Capability Analysis .

Pill Quality Example First we need to do an Attribute Gage R&R • 3 Operators • A few pills • 2 tests/pill Criteria for “good” pill • Logo is clear • No chips or dings • Color applied evenly .

Pill Quality Example Now to collect our data Everyone open your pill bottle Without ingesting the medication: • Count the number of pills • Count how many are defective .

Pill Quality Example Now we will: • Look at the data • Evaluate stability • Determine the distribution • Perform Capability Analysis .

Circuit Board Example We work for an electronics company. and circuit board quality is critical Need to evaluate our capability Remember: • • • • • • Verify measurement system Collect data Look at the data Evaluate stability using a Control Chart Determine the distribution Perform a Capability Analysis .

Circuit Board Example Perform Gage R&R • Approximated as continuous Our criteria: • How many “burn marks” are on the board We will use • 3 operators • 6 parts • 2 runs per part .

record: • How many burn marks there are Destructive test .Circuit Board Example Now everyone please take on circuit board Again without ingesting.

Recap Verify Measurement System • Error types • Ability to measure accurately • “Continuous” versus “Attribute” Tools • Gage R&R • Attribute Gage R&R .

Recap Collect data • Short-term and long-term • Subgrouping • Randomize collection of data Look at the data • Histograms • Boxplots • Other graphs .

Recap Determine distribution of data • • • • Most data is not normal Good fit is critical Some data have natural distribution Probability Plots Evaluate stability • Unstable process is unpredictive • Control Charts .

Recap Perform capability analysis • Capability quantified using several statistics • Capability Analysis Consider data type • Continuous • Binary • Count .

Inc.The End… Joel Smith Commercial Sales Minitab. .