Pn Sample mean: x = n1 i=1 xi Two-tailed Upper-Tailed Lower-Tailed Sample median: x(1) , . . . , x(n) is the ordered sample Hypotheses: H0 : = 0 H0 : = 0 H0 : = 0 x +x Q2 = x(k+1) if n = 2k + 1 and (k) 2 (k+1) if n = 2k H1 : 6= 0 H1 : > 0 H1 : < 0 First quartile: X 0 Q1 divides the bottom 25% of the sample from the top 75% Test Statistic (z-test): Z0 = / n Third quartile: Two-tailed Upper-Tailed Lower-Tailed Q3 divides the bottom 75% of the sample from the top 25% Critical Regions: |z0 | > z/2 z0 > z z0 < z Interquartile range: IQR = Q3 Q1 Two-tailed Upper-Tailed Lower-Tailed Sample mode: value occurring most frequently in data. p-values: 2[1 (|z0 |)] 1 (z0 ) (z0 ) Sample range: R = x(n) x(1) Sample variance: s2 = n1 1 Pn )2 X 0 i=1 (xi x Test Statistic (t-test): T0 = Graphical representation of data: Dot plot, boxplot, Pareto S/ n chart, steam-and-leaf diagram, normplot... Two-tailed Upper-Tailed Lower-Tailed Critical Regions: |t0 | > t/2,n1 t0 > t,n1 t0 < t,n1 Probability Distributions: Two-tailed Upper-Tailed Lower-Tailed E(aX + b) = aE(X) + b and V ar(aX + b) = a2 V ar(X). p-values: 2[1 F (|t0 |)] 1 F (t0 ) F (t0 ) Binomial distribution: X bino(n, p) with E(X) = np and V ar(X) = np(1 p) Type I error: is the error of rejecting H0 when it is true. Poisson distribution: Type II error: is the error of accepting H0 when it is false. X poiss() with E(X) = V ar(X) = Control Charts: Uniform distribution: Western Electric rules: (ba)2 X unif (a, b) with E(X) = a+b 2 and V ar(X) = 12 Rule 1: A point falls outside the upper and lower control Normal distribution: limits, i.e. above U CL or below LCL X N (, 2 ) with E(X) = and V ar(X) = 2 Rule 2: Two out of three consecutive points fall above Standard normal distribution: Z = X N (0, 1) + 2 or below 2 z/2 = 1 1 2 . For = 0.05, z/2 = 1.96
Rule 3: Four out of five consecutive points fall above Normal approximation to binomial: + 1 or below 1 x+0.5np
P (X x) . Rule 4: Eight or more consecutive points fall above or np(1p) Exponential distribution: below X exp() with E(X) = 1 and V ar(X) = 12 Rule 5: Eight or more consecutive points move upward (in- Chi-Squared distribution: creasing) or downward (decreasing) in value. X 2 (n) with E(X) = n and V ar(X) = 2n Control Chart for the Mean (X-chart): Students t-distribution: U CL = x + A2 r CL = x LCL = x A2 r n X t(n) with E(X) = 0 and V ar(X) = n2 Control Chart for the Range (R-chart): F -distribution: X F (n1 , n2 ). U CL = D4 r CL = r LCL = D3 r Control Chart for the Standard Deviation (S-chart): Sampling Distributions: X N (0, 1) U CL = B4 s CL = s LCL = B3 s Z = / n 2 Control Chart for the Mean (from s): (n 1) S2 2 (n 1) + A3 s A3 s U CL = x CL = x LCL = x X
t= S/ n t(n 1) R S Unbiased estimators of are: = d2 or = c4 S 2 / 2 F = SX2 /X2 F (n1 1, n2 1) Y Y Process Capability Analysis: Confidence Intervals: Process capability potential: Cp = U SLLSL 6
100(1 )% confidence interval on with known variance: Upper capability index: CpU = (U SL X)/(3 ) z/2 n x x + z/2 n Lower capability index: CpL = (X LSL)/(3 ) 100(1 )% confidence interval on with unknown variance: Process capability index: Cpk = min(CpU , CpL ) t/2,n1 sn x x + t/2,n1 sn Fraction non-conforming: p = 1 P (LSL < X < U SL). 100(1 )% confidence interval on 2 : (n1)s2 2 Case 1: If Cpk < 1, the process in not capable. 2 2(n1)s 2 /2,n1 1/2,n1 Case 2: If Cpk 1.33, the process in highly capable. 100(1 p)% confidence interval on p:p Case 3: If 1 Cpk < 1.33, the process in barely capable. p z/2 p(1 p)/n p p + z/2 p(1 p)/n