This package aims on Clinical and Laboratory Standards Institute (CLSI) EP15-A3 Calculations
CLSI EP15-A3 provides guidance on the user verification of precision and the estimation of bias for laboratory test methods. It outlines the steps and procedures that clinical laboratories should follow to evaluate the performance of a test method they intend to implement.
This package is a R implementation of the calculations used in the document
Create a table in the specified format
Wider Format
| rep | Run_1 | Run_2 | Run_3 | Run_4 | Run_5 |
|---|---|---|---|---|---|
| 1 | 140 | 140 | 140 | 141 | 139 |
| 2 | 139 | 143 | 138 | 144 | 140 |
| 3 | 138 | 141 | 136 | 142 | 141 |
| 4 | 138 | 143 | 141 | 143 | 138 |
| 5 | 140 | 137 | 136 | 144 | 141 |
Long Format
| rep | name | value |
|---|---|---|
| 1 | Run_1 | 140 |
| 1 | Run_2 | 140 |
| 1 | Run_3 | 140 |
| 1 | Run_4 | 141 |
| 1 | Run_5 | 139 |
| 2 | Run_1 | 139 |
| 2 | Run_2 | 143 |
| 2 | Run_3 | 138 |
| 2 | Run_4 | 144 |
| 2 | Run_5 | 140 |
| 3 | Run_1 | 138 |
| 3 | Run_2 | 141 |
ferritin_long and ferritin_wider are
provided as data in the package and can be used as example
For Long Format
For Wide Format
Calculate Anova parameters and Imprecision Estimates
aov_t <- calculate_aov_infos(data)
aov_t
#> $N
#> [1] 25
#>
#> $k
#> [1] 5
#>
#> $mean
#> [1] 140.12
#>
#> $sd
#> [1] 2.2971
#>
#> $aov_table
#> source_of_variation ss df ms
#> 1 between_run 63.44 4 15.86
#> 2 within_run 63.20 20 3.16
#>
#> $n0
#> [1] 5
#>
#> $Vbetween
#> [1] 2.54
#>
#> $Vwithin
#> [1] 3.16
#>
#> $SR
#> [1] 1.777639
#>
#> $SWL
#> [1] 2.387467
#>
#> $CVR
#> [1] 1.268655
#>
#> $CVWL
#> [1] 1.703873If user repetibility(SR or CVR) < repetibility claim and Within-lab(SWL or CVWL) < Within-lab claim the user has verified manufacture’s precision claims if not the upper verification limit (UVL) should be checked
uvl_info <- calculate_uvl_info(aov_return = aov_t, cvr_or_sr = .43, cvwl_or_swl = .7)
uvl_info
#> $dfR
#> [1] 20
#>
#> $dfWL
#> [1] 8
#>
#> $f_r
#> [1] 1.253205
#>
#> $f_wl
#> [1] 1.392269
#>
#> $cv_uvl_r
#> [1] 0.538878
#>
#> $cv_uvl_wl
#> [1] 0.9745886Where arguments are the follow:
aov_return: Return of calculate_aov_info()
nsamp: number of samples in the experiment. Default is 1
cvr_or_sr: Desirable CV or S repetability
cvwl_or_swl: Desirable CV or S within-lab
Recheck If user repetability(SR or CVR) < UVL repetability claim and Within-lab(SWL or CVWL) < UVL Within-lab claim
For calculating a range for acceptable bias different scenarios and subscenarios are provided by the document
calculate_bias_interval is the function used:
These are the mandatory parameters:
scenario: Choosed scenario from section 3.3 of EP15-A3
nrun: Number of runs
nrep: number of repetitions per run (n0)
SWL: S within laboratory (obtained from anova)
SR: S repetability (obtained from anova)
nsamples: total number of samples tested usual 1
expected_mean: Expected mean or TV
user_mean: Mean of all samples (obtained from anova)
… : additional parameters necessary for processing the choosed scenario
Scenario A
Bona fide reference materials, can vary depending on the information provided by the manufacturer. - Sub scenario “u”: - manufacturer supplies a “standard error,” “standard uncertainty” (u), or “combined standard uncertainty” (often denoted as uC ) for the TV - Sub scenario “Uk”: - manufacturer provides an “expanded uncertainty” (U) for the TV and a “coverage factor” (k) - Sub scenario “Ucoverage”: - manufacturer provides an “expanded uncertainty” (U) for the TV and a “coverage percentage” - Sub scenario “lowerupper”: - manufacturer provides an lower and upper limits and a “coverage percentage” (CI)
Example
calculate_bias_interval('A',
subscenario = 'Uk',
nrun = 7,
nrep = 5,
SWL = .042,
SR = .032,
nsamples = 2,
exppected_mean = 1,
user_mean = .94
)Will return
Error in calculate_se_rm(scenario, additional_args) : For the choosed scenario U and k must be supplied
So we need to pass the requested parameters:
calculate_bias_interval('A',
subscenario = 'Uk',
nrun = 7,
nrep = 5,
SWL = .042,
SR = .032,
nsamples = 2,
expected_mean = 1,
user_mean = .94,
U = 140,
k = 1.96
)
#> $mean
#> [1] 0.94
#>
#> $TV
#> [1] 1
#>
#> $interval
#> $interval$lower_limit
#> [1] -159
#>
#> $interval$higher_limit
#> [1] 161
#>
#>
#> $signif
#> [1] FALSE
#>
#> $bias
#> [1] -0.06Scenario B and C
When a reference material’s total uncertainty (TV) is determined based on Proficiency Testing (PT) (B) or peer group results from an interlaboratory QC program (C)
Additional parameters necessary are sd_rm and nlab
Example
calculate_bias_interval('C', nrun = 7,
nrep = 5,
SWL = .042,
SR = .032,
nsamples = 2,
expected_mean = 1,
user_mean = .94,
sd_rm = .05,
nlab = 43)
#> $mean
#> [1] 0.94
#>
#> $TV
#> [1] 1
#>
#> $interval
#> $interval$lower_limit
#> [1] 0.9684359
#>
#> $interval$higher_limit
#> [1] 1.031564
#>
#>
#> $signif
#> [1] TRUE
#>
#> $bias
#> [1] -0.06Scenario D and E
If the TV represents a conventional quantity value (D) or When working with a commercial QC material supplied with a TV for which the standard error cannot be estimated (E)
calculate_bias_interval('E',
nrun = 7,
nrep = 5,
SWL = .042,
SR = .032,
nsamples = 2,
expected_mean = 1,
user_mean = .94
)
#> $mean
#> [1] 0.94
#>
#> $TV
#> [1] 1
#>
#> $interval
#> $interval$lower_limit
#> [1] 0.9703
#>
#> $interval$higher_limit
#> [1] 1.0297
#>
#>
#> $signif
#> [1] TRUE
#>
#> $bias
#> [1] -0.06Bias conclusion
If the mean is inside interval object returned in
calculate_bias_interval() the result is not significant and
the observed bias is inside the manufacture claims
Package repository: https://github.com/clauciorank/CLSIEP15