Estimates subject-level false-recent rate (FRR) for a given time cutoff. Each subject with any observations after the time cutoff is assigned a recency status according to the majority of observations for that subject after the cutoff. In the event of exactly half of the observations being classified as recent, the subject contributes a count of 0.5. The function performs an exact binomial test and reports the estimated probability of testing recent after the cutoff, a confidence interval for the proportion, the number of recent results ('successes'), number of subjects ('trials') and the number of data points contributing to the subject-level estimate.
frrcal( data = NULL, subid_var = NULL, time_var = NULL, recency_cutoff_time = 730.5, recency_rule = "binary_data", recency_vars = NULL, recency_params = NULL, alpha = 0.05, method = "exact", debug = FALSE )
| data | A data frame containing variables for subject identifier, time (since detectable infection), and variables with biomarker readings or recency status (to be specified in recency_vars) |
|---|---|
| subid_var | The variable in the dataframe identifying subjects |
| time_var | The variable in the dataframe indicating time between 'time zero' (usually detectable infection) and biomarker measurement |
| recency_cutoff_time | Recency time cut-off ('Big T'). Default=730.5. |
| recency_rule | Specified rule for defining recent/non-recent outcomes from biomarker data (see Details) |
| recency_vars | Variables to be used in determining recency outcomes |
| recency_params | Vector of numeric parameters (e.g. thresholds) for determining recency according to the relevant rule |
| alpha | Confidence level, default=0.05. |
| method | Method for computing confidence interval on binomial probability (passed to binom::binom.confint). Default is Clopper-Pearson 'exact' method. Accepted values: `c("exact", "ac", "asymptotic", "wilson", "prop.test", "bayes", "logit", "cloglog", "probit")`. |
| debug | Enable debugging mode (browser) |
The package contains long form documentation in the form of vignettes that cover the use of the main fucntions. Use browseVignettes(package="inctools") to access them.
recency_rule: binary_data - supply a binary variable with 1=recent and 0=non-recent in recency_vars.
recency_rule: independent_thresholds: supply one threshold variable per biomarker in recency_vars and the relevant thresholds, as well as whether a value below or above each threshold indicates recency in recency_params.
recency_params expects a list of pairs of thresholds and thresholdtypes, with zero indicating a reading below the threshold implies recency and 1 that a reading above the threshold implies recency. (Note: two values, a threshold and a thresholdtype per variable must be specified in recency_params. For example, if you specify recency_vars = c('ODn','ViralLoad') you may specify recency_params = c(1.5,0,500,1), meaning that an ODn reading below 1.5 AND a viral load reasing above 500 indicates a recent result. Objects with missing values in its biomarker readings will be excluded from caculation.
frrcal(data=excalibdata, subid_var = "SubjectID", time_var = "DaysSinceEDDI", recency_cutoff_time = 730.5, recency_rule = "independent_thresholds", recency_vars = c("Result","VL"), recency_params = c(10,0,1000,1), method = "exact", alpha = 0.05)#> # A tibble: 1 x 9 #> FRRest SE CI_LB CI_UB alpha n_recent n_subjects #> <dbl> <dbl> <dbl> <dbl> <dbl> <int> <int> #> 1 0.0300752 0.0104721 0.0130720 0.0583978 0.05 8 266 #> # … with 2 more variables: n_observations <int>, ci_method <chr>