Common use of Statistical Methodology Clause in Contracts

Statistical Methodology. 4.1 For all submeasurements with benchmark standards (“benchmark submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing performance levels reported for submeasurements against benchmarks established in the PIDs on a “stare- and-compare” basis (i.e., with no additional statistical methodology applied). 4.2 For all submeasurements with parity standards (“parity submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing statistical z-scores associated with performance levels reported for submeasurements against statistical critical values as defined in Section 5.0. The calculation of z-scores will be based on a statistical test, called the “modified z- test,” as defined in Section 4.4 below, to determine whether a parity condition exists between the results for CenturyLink QC and for CLEC. 4.3 For the purpose of this Section, the CenturyLink QC results will be the CenturyLink QC monthly retail results as specified in the PIDs. 4.4 The modified z-test shall be applicable if the CLEC sample size is greater than 30 for a given submeasurement. The formula for determining parity using the z-test is: z = DIFF / σDIFF Where: DIFF = MCenturyLink QC – MCLEC MCenturyLink QC = CenturyLink QC average or proportion MCLEC = CLEC average or proportion σDIFF = square root [σ2 CenturyLink QC (1/ n CLEC + 1/ n CenturyLink QC)] σ2CenturyLink QC = Calculated variance for CenturyLink QC nCenturyLink QC = number of observations or samples used in CenturyLink QC submeasurement nCLEC = number of observations or samples used in CLEC submeasurement In calculating the difference between CenturyLink QC and CLEC performance, the above formula applies when a larger CenturyLink QC value indicates a better level of performance. In cases where a smaller CenturyLink QC value indicates a higher level of performance, the order is reversed, i.e., MCLEC - MCenturyLink QC . 4.5 For parity submeasurements for which the number of data points is less than or equal to 30, CenturyLink QC will apply a permutation test to determine statistical significance. For such parity submeasurements reported as percentages, where the number of data points is less than or equal to 30, CenturyLink QC will apply an exact proportions test (a form of permutation testing that applies to metrics reported as percentages). The permutation test for metrics reported as intervals will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the z statistic for the actual arrangement of the data. • Pool and mix the CLEC and CenturyLink QC data sets. • Perform the following 1000 times: − Randomly subdivide the pooled data sets into two pools, one the same size as the original CLEC data set (nCLEC) and one reflecting the remaining data points, which is equal to the size of the original CenturyLink QC data set or nCenturyLink QC. − Compute and store the z-test score (ZS) for this sample. • Count the number of times the z statistic for a permutation of the randomly subdivided data is greater than the actual z statistic. • Compute the fraction (p-value) of permutations for which the z statistic for the rearranged data is greater than the z statistic for the actual samples. The exact proportions permutation test for metrics reported as percentages will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the combined (CLEC and Retail) percentage result for the metric. • Identify the possible configurations of Retail metric results and CLEC metric results that could exist in the actual data and yield more extreme differences between CLEC and Retail results, while still yielding the same combined CLEC- Retail result. • For each such configuration of results that yields a more extreme difference than seen in the actual reported results, calculate the probability of observing that more-extreme result, given the actual combined result. • Calculate the sum of the probabilities of the more-extreme data configurations. This sum constitutes the p-value that represents the total probability of observing a more extreme difference between CLEC and Retail results than seen in the actual data. If the resulting p-value is greater than α (alpha), the significance level of the test, the hypothesis of no difference is not rejected, and the test is passed. Alpha = 0.05, except as specified elsewhere herein. For individual month testing for performance measurements involving LIS trunks and DS-1s that are Unbundled Loops (performance measurements: ▇▇-▇▇/▇, ▇▇-▇▇/▇, ▇▇-▇, ▇▇-▇▇/▇, MR-7D/E, and MR-8) with sample sizes of 1-10,

Statistical Methodology. 4.1 For all submeasurements with benchmark standards (“benchmark submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing performance levels reported for submeasurements against benchmarks established in the PIDs on a “stare- and-compare” basis (i.e., with no additional statistical methodology applied). 4.2 For all submeasurements with parity standards (“parity submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing statistical z-scores associated with performance levels reported for submeasurements against statistical critical values as defined in Section 5.0. The calculation of z-scores will be based on a statistical test, called the “modified z- test,” as defined in Section 4.4 below, to determine whether a parity condition exists between the results for CenturyLink QC and for CLEC. 4.3 For the purpose of this Section, the CenturyLink QC results will be the CenturyLink QC monthly retail results as specified in the PIDs. 4.4 The modified z-test shall be applicable if the CLEC sample size is greater than 30 for a given submeasurement. The formula for determining parity using the z-test is: z = DIFF / σDIFF Where: DIFF = MCenturyLink QC – MCLEC MCenturyLink QC = CenturyLink QC average or proportion MCLEC = CLEC average or proportion σDIFF = square root [σ2 CenturyLink QC (1/ n CLEC + 1/ n CenturyLink QC)] σ2CenturyLink CenturyLink QC σ2 = Calculated variance for CenturyLink QC nCenturyLink QC = number of observations or samples used in CenturyLink QC submeasurement nCLEC = number of observations or samples used in CLEC submeasurement In calculating the difference between CenturyLink QC and CLEC performance, the above formula applies when a larger CenturyLink QC value indicates a better level of performance. In cases where a smaller CenturyLink QC value indicates a higher level of performance, the order is reversed, i.e., MCLEC - MCenturyLink QC . 4.5 For parity submeasurements for which the number of data points is less than or equal to 30, CenturyLink QC will apply a permutation test to determine statistical significance. For such parity submeasurements reported as percentages, where the number of data points is less than or equal to 30, CenturyLink QC will apply an exact proportions test (a form of permutation testing that applies to metrics reported as percentages). The permutation test for metrics reported as intervals will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the z statistic for the actual arrangement of the data. • Pool and mix the CLEC and CenturyLink QC data sets. • Perform the following 1000 times: − Randomly subdivide the pooled data sets into two pools, one the same size as the original CLEC data set (nCLEC) and one reflecting the remaining data points, which is equal to the size of the original CenturyLink QC data set or nCenturyLink QC. − Compute and store the z-test score (ZS) for this sample. • Count the number of times the z statistic for a permutation of the randomly subdivided data is greater than the actual z statistic. • Compute the fraction (p-value) of permutations for which the z statistic for the rearranged data is greater than the z statistic for the actual samples. The exact proportions permutation test for metrics reported as percentages will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the combined (CLEC and Retail) percentage result for the metric. • Identify the possible configurations of Retail metric results and CLEC metric results that could exist in the actual data and yield more extreme differences between CLEC and Retail results, while still yielding the same combined CLEC- Retail result. • For each such configuration of results that yields a more extreme difference than seen in the actual reported results, calculate the probability of observing that more-extreme result, given the actual combined result. • Calculate the sum of the probabilities of the more-extreme data configurations. This sum constitutes the p-value that represents the total probability of observing a more extreme difference between CLEC and Retail results than seen in the actual data. If the resulting p-value is greater than α (alpha), the significance level of the test, the hypothesis of no difference is not rejected, and the test is passed. Alpha = 0.05, except as specified elsewhere herein. For individual month testing for performance measurements involving LIS trunks and DS-1s that are Unbundled Loops (performance measurements: ▇▇-▇▇/▇, ▇▇-▇▇/▇, ▇▇-▇, ▇▇-▇▇/▇, MR-7D/E, and MR-8) with sample sizes of 1-10,

Statistical Methodology. 4.1 For all submeasurements with benchmark standards (“benchmark submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing performance levels reported for submeasurements against benchmarks established in the PIDs on a “stare- and-compare” basis (i.e., with no additional statistical methodology applied). 4.2 For all submeasurements with parity standards (“parity submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing statistical z-scores associated with performance levels reported for submeasurements against statistical critical values as defined in Section 5.0. The calculation of z-scores will be based on a statistical test, called the “modified z- test,” as defined in Section 4.4 below, to determine whether a parity condition exists between the results for CenturyLink QC and for CLEC. 4.3 For the purpose of this Section, the CenturyLink QC results will be the CenturyLink QC monthly retail results as specified in the PIDs. 4.4 The modified z-test shall be applicable if the CLEC sample size is greater than 30 for a given submeasurement. The formula for determining parity using the z-test is: z = DIFF / σDIFF Where: DIFF = MCenturyLink QC – MCLEC MCenturyLink QC = CenturyLink QC average or proportion MCLEC = CLEC average or proportion σDIFF = square root [σ2 CenturyLink QC (1/ n CLEC + 1/ n CenturyLink QC)] σ2CenturyLink QC = Calculated variance for CenturyLink QC nCenturyLink QC = number of observations or samples used in CenturyLink QC submeasurement nCLEC = number of observations or samples used in CLEC submeasurement In calculating the difference between CenturyLink QC and CLEC performance, the above formula applies when a larger CenturyLink QC value indicates a better level of performance. In cases where a smaller CenturyLink QC value indicates a higher level of performance, the order is reversed, i.e., MCLEC - MCenturyLink QC . 4.5 For parity submeasurements for which the number of data points is less than or equal to 30, CenturyLink QC will apply a permutation test to determine statistical significance. For such parity submeasurements reported as percentages, where the number of data points is less than or equal to 30, CenturyLink QC will apply an exact proportions test (a form of permutation testing that applies to metrics reported as percentages). The permutation test for metrics reported as intervals will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the z statistic for the actual arrangement of the data. • Pool and mix the CLEC and CenturyLink QC data sets. • Perform the following 1000 times: − Randomly subdivide the pooled data sets into two pools, one the same size as the original CLEC data set (nCLEC) and one reflecting the remaining data points, which is equal to the size of the original CenturyLink QC data set or nCenturyLink QC. − Compute and store the z-test score (ZS) for this sample. • Count the number of times the z statistic for a permutation of the randomly subdivided data is greater than the actual z statistic. • Compute the fraction (p-value) of permutations for which the z statistic for the rearranged data is greater than the z statistic for the actual samples. The exact proportions permutation test for metrics reported as percentages will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the combined (CLEC and Retail) percentage result for the metric. • Identify the possible configurations of Retail metric results and CLEC metric results that could exist in the actual data and yield more extreme differences between CLEC and Retail results, while still yielding the same combined CLEC- Retail result. • For each such configuration of results that yields a more extreme difference than seen in the actual reported results, calculate the probability of observing that more-extreme result, given the actual combined result. • Calculate the sum of the probabilities of the more-extreme data configurations. This sum constitutes the p-value that represents the total probability of observing a more extreme difference between CLEC and Retail results than seen in the actual data. If the resulting p-value is greater than α (alpha), the significance level of the test, the hypothesis of no difference is not rejected, and the test is passed. Alpha = 0.05, except as specified elsewhere herein. For individual month testing for performance measurements involving LIS trunks and DS-1s that are Unbundled Loops (performance measurements: ▇▇-▇▇OP-3D/▇E, ▇▇-▇▇OP-4D/▇E, ▇▇-▇OP-5, ▇▇-▇▇MR-5A/▇B, MR-7D/E, and MR-8) with sample sizes of 1-10,

Statistical Methodology. 4.1 For all submeasurements with benchmark standards (“benchmark submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing performance levels reported for submeasurements against benchmarks established in the PIDs on a “stare- and-compare” basis (i.e., with no additional statistical methodology applied). 4.2 For all submeasurements with parity standards (“parity submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing statistical z-scores associated with performance levels reported for submeasurements against statistical critical values as defined in Section 5.0. The calculation of z-scores will be based on a statistical test, called the “modified z- test,” as defined in Section 4.4 below, to determine whether a parity condition exists between the results for CenturyLink QC and for CLEC. 4.3 For the purpose of this Section, the CenturyLink QC results will be the CenturyLink QC monthly retail results as specified in the PIDs. 4.4 The modified z-test shall be applicable if the CLEC sample size is greater than 30 for a given submeasurement. The formula for determining parity using the z-test is: z = DIFF / σDIFF Where: DIFF = MCenturyLink QC – MCLEC MCenturyLink QC = CenturyLink QC average or proportion MCLEC = CLEC average or proportion σDIFF = square root [σ2 CenturyLink QC (1/ n CLEC + 1/ n CenturyLink QC)] σ2CenturyLink CenturyLink QC σ2 = Calculated variance for CenturyLink QC nCenturyLink QC = number of observations or samples used in CenturyLink QC submeasurement nCLEC = number of observations or samples used in CLEC submeasurement In calculating the difference between CenturyLink QC and CLEC performance, the above formula applies when a larger CenturyLink QC value indicates a better level of performance. In cases where a smaller CenturyLink QC value indicates a higher level of performance, the order is reversed, i.e., MCLEC - MCenturyLink QC . 4.5 For parity submeasurements for which the number of data points is less than or equal to 30, CenturyLink QC will apply a permutation test to determine statistical significance. For such parity submeasurements reported as percentages, where the number of data points is less than or equal to 30, CenturyLink QC will apply an exact proportions test (a form of permutation testing that applies to metrics reported as percentages). The permutation test for metrics reported as intervals will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the z statistic for the actual arrangement of the data. • Pool and mix the CLEC and CenturyLink QC data sets. • Perform the following 1000 times: − Randomly subdivide the pooled data sets into two pools, one the same size as the original CLEC data set (nCLEC) and one reflecting the remaining data points, which is equal to the size of the original CenturyLink QC data set or nCenturyLink QC. − Compute and store the z-test score (ZS) for this sample. • Count the number of times the z statistic for a permutation of the randomly subdivided data is greater than the actual z statistic. • Compute the fraction (p-value) of permutations for which the z statistic for the rearranged data is greater than the z statistic for the actual samples. The exact proportions permutation test for metrics reported as percentages will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the combined (CLEC and Retail) percentage result for the metric. • Identify the possible configurations of Retail metric results and CLEC metric results that could exist in the actual data and yield more extreme differences between CLEC and Retail results, while still yielding the same combined CLEC- Retail result. • For each such configuration of results that yields a more extreme difference than seen in the actual reported results, calculate the probability of observing that more-extreme result, given the actual combined result. • Calculate the sum of the probabilities of the more-extreme data configurations. This sum constitutes the p-value that represents the total probability of observing a more extreme difference between CLEC and Retail results than seen in the actual data. If the resulting p-value is greater than α (alpha), the significance level of the test, the hypothesis of no difference is not rejected, and the test is passed. Alpha = 0.05, except as specified elsewhere herein. For individual month testing for performance measurements involving LIS trunks and DS-1s that are Unbundled Loops (performance measurements: ▇▇-▇▇OP-3D/▇E, ▇▇-▇▇OP-4D/▇E, ▇▇-▇OP-5, ▇▇-▇▇MR-5A/▇B, MR-7D/E, and MR-8) with sample sizes of 1-10,

Statistical Methodology. 4.1 For all submeasurements with benchmark standards (“benchmark submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing performance levels reported for submeasurements against benchmarks established in the PIDs on a “stare- and-compare” basis (i.e., with no additional statistical methodology applied). 4.2 For all submeasurements with parity standards (“parity submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing statistical z-scores associated with performance levels reported for submeasurements against statistical critical values as defined in Section 5.0. The calculation of z-scores will be based on a statistical test, called the “modified z- test,” as defined in Section 4.4 below, to determine whether a parity condition exists between the results for CenturyLink QC and for CLEC. 4.3 For the purpose of this Section, the CenturyLink QC results will be the CenturyLink QC monthly retail results as specified in the PIDs. 4.4 The modified z-test shall be applicable if the CLEC sample size is greater than 30 for a given submeasurement. The formula for determining parity using the z-test is: z = DIFF / σDIFF DIFF Where: DIFF = MCenturyLink QC – MCLEC MCenturyLink QC = CenturyLink QC average or proportion MCLEC = CLEC average or proportion σDIFF DIFF = square root [σ2 2 CenturyLink QC (1/ n CLEC + 1/ n CenturyLink QC▇/ ▇ ▇▇▇▇▇▇▇▇▇▇▇ ▇▇)] σ2CenturyLink QC = Calculated variance for CenturyLink QC nCenturyLink QC = number of observations or samples used in CenturyLink QC submeasurement nCLEC = number of observations or samples used in CLEC submeasurement In calculating the difference between CenturyLink QC and CLEC performance, the above formula applies when a larger CenturyLink QC value indicates a better level of performance. In cases where a smaller CenturyLink QC value indicates a higher level of performance, the order is reversed, i.e., MCLEC - MCenturyLink QC . 4.5 For parity submeasurements for which the number of data points is less than or equal to 30, CenturyLink QC will apply a permutation test to determine statistical significance. For such parity submeasurements reported as percentages, where the number of data points is less than or equal to 30, CenturyLink QC will apply an exact proportions test (a form of permutation testing that applies to metrics reported as percentages). The permutation test for metrics reported as intervals will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the z statistic for the actual arrangement of the data. • Pool and mix the CLEC and CenturyLink QC data sets. • Perform the following 1000 times: − Randomly subdivide the pooled data sets into two pools, one the same size as the original CLEC data set (nCLEC) and one reflecting the remaining data points, which is equal to the size of the original CenturyLink QC data set or nCenturyLink QC. − Compute and store the z-test score (ZS) for this sample. • Count the number of times the z statistic for a permutation of the randomly subdivided data is greater than the actual z statistic. • Compute the fraction (p-value) of permutations for which the z statistic for the rearranged data is greater than the z statistic for the actual samples. The exact proportions permutation test for metrics reported as percentages will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the combined (CLEC and Retail) percentage result for the metric. • Identify the possible configurations of Retail metric results and CLEC metric results that could exist in the actual data and yield more extreme differences between CLEC and Retail results, while still yielding the same combined CLEC- Retail result. • For each such configuration of results that yields a more extreme difference than seen in the actual reported results, calculate the probability of observing that more-extreme result, given the actual combined result. • Calculate the sum of the probabilities of the more-extreme data configurations. This sum constitutes the p-value that represents the total probability of observing a more extreme difference between CLEC and Retail results than seen in the actual data. If the resulting p-value is greater than α (alpha), the significance level of the test, the hypothesis of no difference is not rejected, and the test is passed. Alpha = 0.05, except as specified elsewhere herein. For individual month testing for performance measurements involving LIS trunks and DS-1s that are Unbundled Loops (performance measurements: ▇▇-▇▇/▇, ▇▇-▇▇/▇, ▇▇-▇, ▇ ▇▇-▇▇/▇, MR-7D/E, and MR-8) with sample sizes of 1-10,

Statistical Methodology. 4.1 For all submeasurements with benchmark standards (“benchmark submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing performance levels reported for submeasurements against benchmarks established in the PIDs on a “stare- and-compare” basis (i.e., with no additional statistical methodology applied). 4.2 For all submeasurements with parity standards (“parity submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing statistical z-scores associated with performance levels reported for submeasurements against statistical critical values as defined in Section 5.0. The calculation of z-scores will be based on a statistical test, called the “modified z- test,” as defined in Section 4.4 below, to determine whether a parity condition exists between the results for CenturyLink QC and for CLEC. 4.3 For the purpose of this Section, the CenturyLink QC results will be the CenturyLink QC monthly retail results as specified in the PIDs. 4.4 The modified z-test shall be applicable if the CLEC sample size is greater than 30 for a given submeasurement. The formula for determining parity using the z-test is: z = DIFF / σDIFF DIFF Where: DIFF = MCenturyLink QC – MCLEC MCenturyLink QC = CenturyLink QC average or proportion MCLEC = CLEC average or proportion σDIFF DIFF = square root [σ2 2 CenturyLink QC (1/ n CLEC + 1/ n CenturyLink QC)] σ2CenturyLink 2CenturyLink QC = Calculated variance for CenturyLink QC nCenturyLink QC = number of observations or samples used in CenturyLink QC submeasurement nCLEC = number of observations or samples used in CLEC submeasurement In calculating the difference between CenturyLink QC and CLEC performance, the above formula applies when a larger CenturyLink QC value indicates a better level of performance. In cases where a smaller CenturyLink QC value indicates a higher level of performance, the order is reversed, i.e., MCLEC - MCenturyLink QC . 4.5 For parity submeasurements for which the number of data points is less than or equal to 30, CenturyLink QC will apply a permutation test to determine statistical significance. For such parity submeasurements reported as percentages, where the number of data points is less than or equal to 30, CenturyLink QC will apply an exact proportions test (a form of permutation testing that applies to metrics reported as percentages). The permutation test for metrics reported as intervals will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the z statistic for the actual arrangement of the data. • Pool and mix the CLEC and CenturyLink QC data sets. • Perform the following 1000 times: − Randomly subdivide the pooled data sets into two pools, one the same size as the original CLEC data set (nCLEC) and one reflecting the remaining data points, which is equal to the size of the original CenturyLink QC data set or nCenturyLink QC. − Compute and store the z-test score (ZS) for this sample. • Count the number of times the z statistic for a permutation of the randomly subdivided data is greater than the actual z statistic. • Compute the fraction (p-value) of permutations for which the z statistic for the rearranged data is greater than the z statistic for the actual samples. The exact proportions permutation test for metrics reported as percentages will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the combined (CLEC and Retail) percentage result for the metric. • Identify the possible configurations of Retail metric results and CLEC metric results that could exist in the actual data and yield more extreme differences between CLEC and Retail results, while still yielding the same combined CLEC- Retail result. • For each such configuration of results that yields a more extreme difference than seen in the actual reported results, calculate the probability of observing that more-extreme result, given the actual combined result. • Calculate the sum of the probabilities of the more-extreme data configurations. This sum constitutes the p-value that represents the total probability of observing a more extreme difference between CLEC and Retail results than seen in the actual data. If the resulting p-value is greater than α  (alpha), the significance level of the test, the hypothesis of no difference is not rejected, and the test is passed. Alpha = 0.05, except as specified elsewhere herein. For individual month testing for performance measurements involving LIS trunks and DS-1s that are Unbundled Loops (performance measurements: ▇▇-▇▇/▇, ▇▇-▇▇/▇, ▇▇-▇, ▇▇-▇▇/▇, MR-7D/E, and MR-8) with sample sizes of 1-10,

Statistical Methodology. 4.1 For all submeasurements with benchmark standards (“benchmark submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing performance levels reported for submeasurements against benchmarks established in the PIDs on a “stare- and-compare” basis (i.e., with no additional statistical methodology applied). 4.2 For all submeasurements with parity standards (“parity submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing statistical z-scores associated with performance levels reported for submeasurements against statistical critical values as defined in Section 5.0. The calculation of z-scores will be based on a statistical test, called the “modified z- test,” as defined in Section 4.4 below, to determine whether a parity condition exists between the results for CenturyLink QC and for CLEC. 4.3 For the purpose of this Section, the CenturyLink QC results will be the CenturyLink QC monthly retail results as specified in the PIDs. 4.4 The modified z-test shall be applicable if the CLEC sample size is greater than 30 for a given submeasurement. The formula for determining parity using the z-test is: z = DIFF / σDIFF Where: DIFF = MCenturyLink QC – MCLEC MCenturyLink QC = CenturyLink QC average or proportion MCLEC = CLEC average or proportion σDIFF = square root [σ2 CenturyLink QC (1/ n CLEC + 1/ n CenturyLink QC)] σ2CenturyLink CenturyLink QC σ2 = Calculated variance for CenturyLink QC nCenturyLink QC = number of observations or samples used in CenturyLink QC submeasurement nCLEC = number of observations or samples used in CLEC submeasurement In calculating the difference between CenturyLink QC and CLEC performance, the above formula applies when a larger CenturyLink QC value indicates a better level of performance. In cases where a smaller CenturyLink QC value indicates a higher level of DocuSign Envelope ID: 4647CA25-E106-49CD-BD1B-8FFCA893A3C8 performance, the order is reversed, i.e., MCLEC - MCenturyLink QC .. DocuSign Envelope ID: 4647CA25-E106-49CD-BD1B-8FFCA893A3C8 4.5 For parity submeasurements for which the number of data points is less than or equal to 30, CenturyLink QC will apply a permutation test to determine statistical significance. For such parity submeasurements reported as percentages, where the number of data points is less than or equal to 30, CenturyLink QC will apply an exact proportions test (a form of permutation testing that applies to metrics reported as percentages). The permutation test for metrics reported as intervals will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the z statistic for the actual arrangement of the data. • Pool and mix the CLEC and CenturyLink QC data sets. • Perform the following 1000 times: − Randomly subdivide the pooled data sets into two pools, one the same size as the original CLEC data set (nCLEC) and one reflecting the remaining data points, which is equal to the size of the original CenturyLink QC data set or nCenturyLink QC. − Compute and store the z-test score (ZS) for this sample. • Count the number of times the z statistic for a permutation of the randomly subdivided data is greater than the actual z statistic. • Compute the fraction (p-value) of permutations for which the z statistic for the rearranged data is greater than the z statistic for the actual samples. The exact proportions permutation test for metrics reported as percentages will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the combined (CLEC and Retail) percentage result for the metric. • Identify the possible configurations of Retail metric results and CLEC metric results that could exist in the actual data and yield more extreme differences between CLEC and Retail results, while still yielding the same combined CLEC- Retail result. • For each such configuration of results that yields a more extreme difference than seen in the actual reported results, calculate the probability of observing that more-extreme result, given the actual combined result. • Calculate the sum of the probabilities of the more-extreme data configurations. This sum constitutes the p-value that represents the total probability of observing a more extreme difference between CLEC and Retail results than seen in the actual data. If the resulting p-value is greater than α (alpha), the significance level of the test, the hypothesis of no difference is not rejected, and the test is passed. Alpha = 0.05, except as specified elsewhere herein. For individual month testing for performance measurements involving LIS trunks and DS-1s that are Unbundled Loops (performance measurements: ▇▇DocuSign Envelope ID: 4647CA25-▇▇E106-49CD-BD1B-8FFCA893A3C8 OP-3D/▇E, ▇▇-▇▇OP-4D/▇E, ▇▇-▇OP-5, ▇▇-▇▇MR-5A/▇B, MR-7D/E, and MR-8) with sample sizes of 1-10,, DocuSign Envelope ID: 4647CA25-E106-49CD-BD1B-8FFCA893A3C8

Statistical Methodology. 4.1 For all submeasurements with benchmark standards (“benchmark submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing performance levels reported for submeasurements against benchmarks established in the PIDs on a “stare- and-compare” basis (i.e., with no additional statistical methodology applied). 4.2 For all submeasurements with parity standards (“parity submeasurements”), as designated in the PIDs, the determination of CenturyLink QC’s conformance with Plan and PID standards will involve comparing statistical z-scores associated with performance levels reported for submeasurements against statistical critical values as defined in Section 5.0. The calculation of z-scores will be based on a statistical test, called the “modified z- test,” as defined in Section 4.4 below, to determine whether a parity condition exists between the results for CenturyLink QC and for CLEC. 4.3 For the purpose of this Section, the CenturyLink QC results will be the CenturyLink QC monthly retail results as specified in the PIDs. 4.4 The modified z-test shall be applicable if the CLEC sample size is greater than 30 for a given submeasurement. The formula for determining parity using the z-test is: z = DIFF / σDIFF Where: DIFF = MCenturyLink QC – MCLEC MCenturyLink QC = CenturyLink QC average or proportion MCLEC = CLEC average or proportion σDIFF DIFF = square root [σ2 2 CenturyLink QC (1/ n CLEC + 1/ n CenturyLink QC)] σ2CenturyLink CenturyLink QC 2 = Calculated variance for CenturyLink QC nCenturyLink QC = number of observations or samples used in CenturyLink QC submeasurement nCLEC = number of observations or samples used in CLEC submeasurement In calculating the difference between CenturyLink QC and CLEC performance, the above formula applies when a larger CenturyLink QC value indicates a better level of performance. In cases where a smaller CenturyLink QC value indicates a higher level of performance, the order is reversed, i.e., MCLEC - MCenturyLink QC . 4.5 For parity submeasurements for which the number of data points is less than or equal to 30, CenturyLink QC will apply a permutation test to determine statistical significance. For such parity submeasurements reported as percentages, where the number of data points is less than or equal to 30, CenturyLink QC will apply an exact proportions test (a form of permutation testing that applies to metrics reported as percentages). The permutation test for metrics reported as intervals will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the z statistic for the actual arrangement of the data. • Pool and mix the CLEC and CenturyLink QC data sets. • Perform the following 1000 times: − Randomly subdivide the pooled data sets into two pools, one the same size as the original CLEC data set (nCLEC) and one reflecting the remaining data points, which is equal to the size of the original CenturyLink QC data set or nCenturyLink QC. − Compute and store the z-test score (ZS) for this sample. • Count the number of times the z statistic for a permutation of the randomly subdivided data is greater than the actual z statistic. • Compute the fraction (p-value) of permutations for which the z statistic for the rearranged data is greater than the z statistic for the actual samples. The exact proportions permutation test for metrics reported as percentages will be applied to calculate the z statistic using the following logic or an equivalent approach that would yield the same result: • Calculate the combined (CLEC and Retail) percentage result for the metric. • Identify the possible configurations of Retail metric results and CLEC metric results that could exist in the actual data and yield more extreme differences between CLEC and Retail results, while still yielding the same combined CLEC- Retail result. • For each such configuration of results that yields a more extreme difference than seen in the actual reported results, calculate the probability of observing that more-extreme result, given the actual combined result. • Calculate the sum of the probabilities of the more-extreme data configurations. This sum constitutes the p-value that represents the total probability of observing a more extreme difference between CLEC and Retail results than seen in the actual data. If the resulting p-value is greater than α (alpha), the significance level of the test, the hypothesis of no difference is not rejected, and the test is passed. Alpha = 0.05, except as specified elsewhere herein. For individual month testing for performance measurements involving LIS trunks and DS-1s that are Unbundled Loops (performance measurements: ▇▇-▇▇/▇, ▇▇-▇▇/▇, ▇▇-▇, ▇▇-▇▇/▇, MR-7D/E, and MR-8) with sample sizes of 1-10,