Bioequivalence and Bioavailability Forum • replicateBE: Data import

replicateBE: Data import [🇷 for BE/BA]

posted by Helmut – Vienna, Austria, 2022-09-17 13:02 (752 d 20:59 ago) – Posting: # 23305
Views: 2,894

Hi Thalita,

❝ First, as stated by roman_max, thanks you, Michael and Detlew for all efforts in this package!

Welcome.

❝ So, I am dealing with some troubles when I run the methods A and B in the package to analyse a pivotal study …

❝

❝ I prepare my datafile as descripted on "DataStructure" in help(p=replicateBE) and save as a CSV: subject, period, sequence, treatment, PK, logPK

So far, so good. You don’t need the column logPK. By default the library will use values given in the column PK and transform the data internally in full numeric precision.

❝ Following the example of GitHUB, …

Such an example is not given on GitHub. :confused:

❝ … I am using the code:

❝

❝ cabo <- read.csv(file="Cabo_Outcome.CSV", head=TRUE, sep=",")

❝ cabo

❝ subject period sequence treatment PK logPK

❝ 1 1 1 RTTR R 23017.637 4.362060731

❝ ...

❝ 20 5 4 RTTR R 25284.525 4.402854806

A pivotal study with five subjects? Your data are extremely imbalanced (four subjects in sequence RTTR and one in sequence TRRT). Pretty unusual. If you read the file in 🇷 you get

str(cabo) 'data.frame': 20 obs. of 6 variables: $ subject : int 1 1 1 1 2 2 2 2 3 3 ... $ period : int 1 2 3 4 1 2 3 4 1 2 ... $ sequence : chr "RTTR" "RTTR" "RTTR" "RTTR" ... $ treatment: chr "R" "T" "T" "R" ... $ PK : num 23018 26945 27755 23902 7715 ... $ logPK : num 4.36 4.43 4.44 4.38 3.89 ...

As you see subject, period are integers and sequence, treatment are characters. For the linear model they have to be factors – and many more preparations have to be done before the data can be used.¹

❝ res <- method.A(verbose = TRUE, details = TRUE,

❝ print = FALSE, data = cabo)

❝

❝ Error in if (!act %in% ref) { : argument is of length zero

By design. :-D

You should not use method.A(..., data = cabo). Although that’s possible in principle, I don’t recommend it. Let the library do the import as explained in help(method.A).

library(replicateBE) path <- "your path" res <- method.A(path.in = path, path.out = path, file = "Cabo_Outcome", ext = "CSV", print = FALSE, details = TRUE, verbose = TRUE)

You should get in the console:

Data set Cabo_Outcome: Method A by lm() --------------------------------------- Type III Analysis of Variance Table Response: log(PK) Df Sum Sq Mean Sq F value Pr(>F) sequence 1 0.150001 0.150001 0.15981 0.71609068 period 3 0.954294 0.318098 5.51168 0.01475522 treatment 1 0.094433 0.094433 1.63624 0.22716731 sequence:subject 3 2.815829 0.938610 16.26328 0.00023442 Residuals 11 0.634848 0.057713 treatment T – R: Estimate Std. Error t value Pr(>|t|) -0.171786 0.134296 -1.279150 0.227167 11 Degrees of Freedom

To show the results:

print(res, row.names = FALSE) Design Method n nTT nRR Sub/seq Miss/seq Miss/per alpha DF CVwT(%) TRRT|RTTR A 5 5 5 1|4 0|0 0|0|0|0 0.05 11 33.1855982312 CVwR(%) swT swR sw.ratio sw.ratio.CL 19.7497208298 0.323226975979 0.195611007311 1.65239666429 5.03279665904 BE.lo(%) BE.hi(%) CL.lo(%) CL.hi(%) PE(%) CI GMR BE 80 125 66.1684875056 107.185933986 84.2159790878 fail pass fail log.half-width sw.ratio.rec.CL 0.241180348576 NA

I recommend to print to a file (that’s the default of the functions).

method.A(path.in = path, path.out = path, file = "Cabo_Outcome", ext = "CSV")

Part of Cabo_Outcome_ABEL_MethodA.txt:

R version : 4.2.1 (2022-06-23 ucrt) PowerTOST version : 1.5.4 (2022) replicateBE version: 1.1.3 (2022) ─────────────────────────────────────────────────────────────────────────────── Function : CV.calc(): stats:lm() executed 2022-09-17 11:00:22 CEST Fixed effects : sequence, subject(sequence), period Data : treatment = R Function : method.A(): stats:lm() executed 2022-09-17 11:00:22 CEST Fixed effects : sequence, subject(sequence), period, treatment Data : all ─────────────────────────────────────────────────────────────────────────────── Analysis performed on column ‘PK’ (data internally log-transformed) Sequences (design) : TRRT|RTTR (4-period full replicate) Subjects / sequence: 1|4 (unbalanced) Subjects (total) : 5 Subj’s with T and R: 5 (calculation of the CI) Less than 12 as required acc. to the BE-GL. Subj’s with two Ts : 5 Subj’s with two Rs : 5 Degrees of freedom : 11 alpha : 0.05 (90% CI) Regulator : EMA Switching CV : 30.00% Scaling cap : 50.00% Regul. constant (k): 0.760 CVwT : 33.19% swT : 0.32323 CVwR : 19.75% (reference-scaling not applicable) BE-limits : 80.00% ... 125.00% swT / swR : 1.6524 (T higher variability than R) sw-ratio (upper CL): 5.0328 (T higher variability than R) Confidence interval: 66.17% ... 107.19% fail Point estimate : 84.22% pass Mixed (CI & PE) : fail

❝ … and calculate CVw

The library will not give you CVw.² In your full replicate design you get CVwT and CVwR. It’s a bit tricky if you are really interested in it (works also for a partial replicate design):

path <- "your path" # use the internal function to import and prepare the data x <- replicateBE:::get.data(path.in = path, path.out = path, file = "Cabo_Outcome", set = "", ext = "CSV", print = FALSE, plot.bxp = FALSE, data = NULL) # conventional linear model mod <- lm(log(PK) ~ sequence + subject %in% sequence + period + treatment, data = x$data) mse <- anova(mod)["Residuals", "Mean Sq"] # residual mean squares error CVw <- sqrt(exp(mse) - 1) # convert to within-subject CV cat(sprintf("CVw = %.2f%%\n", 100 * CVw)) CVw = 24.37%

If you fine with copypasting, use the result of the ANOVA directly:

cat(sprintf("CVw = %.2f%%\n", 100 * sqrt(exp(0.057713) - 1))) CVw = 24.37%

Hope that helps. If not, feel free to ask.

The internal function get.data() will do a lot of stuff. Then the data is a list with this structure:
List of 16 $ data :'data.frame': 20 obs. of 6 variables: ..$ subject : Factor w/ 5 levels "1","2","3","4",..: 1 1 1 1 2 2 2 2 3 3 ... ..$ period : Factor w/ 4 levels "1","2","3","4": 1 2 3 4 1 2 3 4 1 2 ... ..$ sequence : Factor w/ 2 levels "RTTR","TRRT": 1 1 1 1 2 2 2 2 1 1 ... ..$ treatment: Factor w/ 2 levels "R","T": 1 2 2 1 2 1 1 2 1 2 ... ..$ PK : num [1:20] 23018 26945 27755 23902 7715 ... ..$ logPK : num [1:20] 4.36 4.43 4.44 4.38 3.89 ... $ ref :'data.frame': 10 obs. of 6 variables: ..$ subject : Factor w/ 5 levels "1","2","3","4",..: 1 1 2 2 3 3 4 4 5 5 ..$ period : Factor w/ 4 levels "1","2","3","4": 1 4 2 3 1 4 1 4 1 4 ..$ sequence : Factor w/ 2 levels "RTTR","TRRT": 1 1 2 2 1 1 1 1 1 1 ..$ treatment: Factor w/ 2 levels "R","T": 1 1 1 1 1 1 1 1 1 1 ..$ PK : num [1:10] 23018 23902 21282 13604 8982 ... ..$ logPK : num [1:10] 4.36 4.38 4.33 4.13 3.95 ... $ RR :'data.frame': 10 obs. of 6 variables: ..$ subject : Factor w/ 5 levels "1","2","3","4",..: 1 1 2 2 3 3 4 4 5 5 ..$ period : Factor w/ 4 levels "1","2","3","4": 1 4 2 3 1 4 1 4 1 4 ..$ sequence : Factor w/ 2 levels "RTTR","TRRT": 1 1 2 2 1 1 1 1 1 1 ..$ treatment: Factor w/ 2 levels "R","T": 1 1 1 1 1 1 1 1 1 1 ..$ PK : num [1:10] 23018 23902 21282 13604 8982 ... ..$ logPK : num [1:10] 4.36 4.38 4.33 4.13 3.95 ... $ test :'data.frame': 10 obs. of 6 variables: ..$ subject : Factor w/ 5 levels "1","2","3","4",..: 1 1 2 2 3 3 4 4 5 5 ..$ period : Factor w/ 4 levels "1","2","3","4": 2 3 1 4 2 3 2 3 2 3 ..$ sequence : Factor w/ 2 levels "RTTR","TRRT": 1 1 2 2 1 1 1 1 1 1 ..$ treatment: Factor w/ 2 levels "R","T": 2 2 2 2 2 2 2 2 2 2 ..$ PK : num [1:10] 26945 27755 7715 11701 21978 ... ..$ logPK : num [1:10] 4.43 4.44 3.89 4.07 4.34 ... $ type : chr "TRRT|RTTR" $ n : int 5 $ nTT : int 5 $ nRR : int 5 $ design : chr "full" $ txt : chr "\nAnalysis performed on column ‘PK’ (data internally log-transformed)\nSequences (design) : TRRT|RTTR (4-period"| __truncated__ $ Sub.Seq : 'table' int [1:2(1d)] 1 4 ..- attr(*, "dimnames")=List of 1 .. ..$ : chr [1:2] "TRRT" "RTTR" $ Miss.seq: Named num [1:2] 0 0 ..- attr(*, "names")= chr [1:2] "TRRT" "RTTR" $ Miss.per: num [1:4] 0 0 0 0 $ logtrans: logi TRUE $ res.file: logi NA $ png.path: logi NA

This explains
❝ Error in if (!act %in% ref) { : argument is of length zero
In cobo you have only the data frame data (not factorized; see above) but the others are missing. The error is caused by the missing data frame ref containing only the data of the reference.
You are the first one to ask for it (the packages had 22,000 downloads so far). If you explain why you need it, I will consider to add it to the next release.

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

Complete thread:

replicateBE_xlsx datafile roman_max 2020-06-10 17:32 [🇷 for BE/BA]
- replicateBE: data input Helmut 2020-06-10 18:14
  - replicateBE: data input roman_max 2020-06-22 17:06
  - replicateBE: error in data input thalita_m 2022-09-17 05:28
    - replicateBE: Data importHelmut 2022-09-17 11:02