Compute AR1 correlation coefficient

Description

Computes value for AR1 correlation coefficient for use in p_adjust

Usage

ar1Solver(t)

Arguments

Value

Estimated AR1 correlation coefficient

See Also

p_adjust

Examples

t <- rt(1e3, df = 1)
rho <- ar1Solver(t)

bdots Remove Function

Description

Remove observations with a specified fitCode and optionally all pairs

Usage

bdRemove(bdObj, fitCode = 6L, removePairs = TRUE)

Arguments

Details

This function is used to remove all bdots observations with a fit code equal to or larger than the argument passed to fitCode without refitting. If removePairs = TRUE, all entries for a subject will be removed if their fit failed in any of the groups in which they were a member

Create bootstrapped curves from bdotsObj

Description

Creates bootstrapped curves and performs alpha adjustment. Can perform "difference of difference" for nested comparisons

Usage

bdotsBoot(
  formula,
  bdObj,
  Niter = 1000,
  alpha = 0.05,
  padj = "oleson",
  cores = 0,
  ...
)

Arguments

Details

The formula is the only tricky part of this. There will be a minor update to how it works in the future. The three parts we will examine here are Groups, the LHS, and the RHS. For all variable names, special characters should be included with backticks, i.e., `my-var`

## Groups

The Groups are the values input in group in the bdotsFit function, which are columns of the dataset used. These will be denoted G_i Within each group, we will designate the unique values within each group as v_j, ..., whereby G_i(v_1, v_2) will designate unique two unique values within G_i. The possible values of v_i will be implied by the group with which they are associated.

For example, if we have groups vehicle and color, we could specify that we are interested in all blue cars and trucks with the expression vehicle(car, truck) + color(red).

## Formula

### Bootstrapped difference of curves

This illustrates the case in which we are taking a simple bootstraped difference between two curves within a single group

If only one group was provided in bdotsFit, we can take the bootstrapped difference between two values within the group with

y ~ Group1(val1, val2)

If more than two groups were provided, we must specify within which values of the other groups we would like to compare the differences from Group1 in order to uniquely identify the observations. This would be

y ~ Group1(val1, val2) + Group2(val1)

For example, bootstrapping the differences between cars and trucks when color was provided as a second group, we would need y ~ vehicle(car, truck) + color(red).

### Bootstrapped difference of difference curves

This next portion illustrates the case in which we are interested in studying the difference between the differences between two groups, which we will call the innerGroup and the outerGroup following a nested container metaphor. Here, we must use caution as the order of these differences matter. Using again the vehicle example, we can describe this in two ways:

1. We may be interested in comparing the difference between red trucks and cars (d_red) with the difference between blue trucks and cars (d_blue). In this case, we will be finding the difference between cars and trucks twice (one for blue, one for red). The vehicle type is the innerGroup, nested within the outerGroup, in this case, color.

2. We may also be interested in comparing the difference between red trucks and blue trucks (d_truck) with the difference between red and blue cars (d_car). Here, innerGroup is the color and outerGroup is the vehicle

As our primary object of interest here is not the difference in outcome itself, but the difference of the outcome within two groups, the LHS of the formula is written diffs(y, Group1(val1, val2)), where Group1 is the innerGroup. The RHS is then used to specify the groups of which we want to take the inner difference of. The syntax here is the same as above. Together, then, the formula looks like

diffs(y, Group1(val1, val2)) ~ Group2(val1, val2)

in the case in which only two grouping variables were provided to bdotsFit and

diffs(y, Group1(val1, val2)) ~ Group2(val1, val2) + Group3(val1) + ...

is used to uniquely identify the sets of differences when three or more groups were provided.

Value

Object of class 'bdotsBootObj'

Examples

## Not run: 

## fit <- bdotsFit(cohort_unrelated, ...)

boot1 <- bdotsBoot(formula = diffs(Fixations, LookType(Cohort, Unrelated_Cohort)) ~ Group(50, 65),
                   bdObj = fit,
                   N.iter = 1000,
                   alpha = 0.05,
                   p.adj = "oleson",
                   cores = 4)

boot2 <- bdotsBoot(formula = Fixations ~ Group(50, 65) + LookType(Cohort),
                   bdObj = fit,
                   N.iter = 1000,
                   alpha = 0.05,
                   p.adj = "oleson",
                   cores = 4)

## End(Not run)

Correlation with fixed value in bdots

Description

Find the correlation of a fixed value with the bdots fitted curves at each time point

Usage

bdotsCorr(bdObj, val, ciBands = FALSE, method = "pearson")

Arguments

Fit nlme curves to grouped observations

Description

Creates observation level curves to use in bdotsBoot

Usage

bdotsFit(
  data,
  subject,
  time,
  y,
  group,
  curveType = doubleGauss(concave = TRUE),
  cor = TRUE,
  numRefits = 0,
  cores = 0,
  verbose = FALSE,
  ...
)

Arguments

Details

This is step one of the three step bdots process. Things should be more or less straight forward. The only tricky part involves curveType. For now know that one can use doubleGauss(concave = TRUE/FALSE) or logistic(). Should be passed in as a call. See the vignette on customizing this

Value

Object of class 'bdotsObj', inherits from data.table

Examples

## Not run: 
res <- bdotsFit(data = cohort_unrelated,
                subject = "Subject",
                time = "Time",
                y = "Fixations",
                group = c("Group", "LookType"),
                curveType = doubleGauss(concave = TRUE),
                cor = TRUE,
                numRefits = 2,
                cores = 0,
                verbose = FALSE)

## End(Not run)

Fits Individual Subject Curve

Description

The one subject version of bdotsFit

Usage

bdotsFitter(
  dat,
  curveType,
  rho,
  numRefits = 0,
  verbose,
  getCovOnly = NULL,
  params = NULL,
  splitVars = NULL,
  datVarNames = NULL,
  ...
)

Arguments

Refit Observations Returned from bdotsFit

Description

Refit Observations Returned from bdotsFit

Usage

bdotsRefit(
  bdObj,
  fitCode = 1L,
  quickRefit = FALSE,
  numRefits = 2L,
  paramDT = NULL,
  ...
)

Arguments

Details

fitCode indicates lower bound on observations to refit. For example, if fitCode = 4, bdotsRefit will prompt user to refit all observations with fitCode = 4, 5, 6. The quickRit option will attempt to jitter and refit all observations selected by fitCode. Otherwise, the user will be prompted through a menu to individually refit observations

Value

Returns bdObj with updated fits

ci dataset

Description

ci dataset - need to include details

Usage

ci

Format

An object of class data.frame with 108216 rows and 5 columns.

Extract bdotsFit Moedel Coefficients

Description

Returns coefficient matrix for bdotsFit object

Usage

## S3 method for class 'bdotsObj'
coef(object, ...)

Arguments

Value

Returns matrix of model coefficients for observations in object

Create data.table with bdotsObj parameters

Description

Creates an object of class data.table that matches parameter values for each observation. This can then be passed to the bdotsRefit function

Usage

coefWriteout(bdObj)

Arguments

Value

A data.table matching parameter values to observations

Examples

## Not run: 
fit <- bdotsFit(data = cohort_unrelated,
                subject = "Subject",
                time = "Time",
                y = "Fixations",
                group = c("Group", "LookType"),
                curveType = doubleGauss(concave = TRUE),
                cor = TRUE,
                numRefits = 2,
                cores = 0,
                verbose = FALSE)
parDT <- coefWriteout(fit)

## End(Not run)

cohort_unrelated dataset

Description

cohort_unrelated dataset - need to include details

Usage

cohort_unrelated

Format

An object of class data.frame with 50100 rows and 6 columns.

Curve Fitter

Description

Used in bdotsFit

Usage

curveFitter(dat, ff, params, rho, numRefits = 0, getCovOnly = NULL, ...)

Arguments

df_cohort_unrelated dataset

Description

df_cohort_unrelated dataset - need to include details

Usage

df_cohort_unrelated

Format

An object of class data.frame with 78156 rows and 5 columns.

df_target dataset

Description

df_target dataset - need to include details

Usage

df_target

Format

An object of class data.frame with 37575 rows and 4 columns.

Double Gauss curve function for nlme

Description

Double Gauss function used in fitting nlme curve for observations

Usage

doubleGauss(dat, y, time, params = NULL, concave = TRUE, ...)

Arguments

Details

User should only have to worry about setting concavity of this function

y ~ (time < mu) * (exp(-1 * (time - mu) ^ 2 / (2 * sig1 ^ 2)) * (ht - base1) + base1) + (mu <= time) * (exp(-1 * (time - mu) ^ 2 / (2 * sig2 ^ 2)) * (ht - base2) + base2)

DoubleGauss2 curve function for nlme

Description

DoubleGauss2 function used in fitting nlme curve for observations

Usage

doubleGauss2(dat, y, time, params = NULL, concave = TRUE, ...)

Arguments

Details

User should only have to worry about setting concavity of this function. Presently only work for time series scaled out to 2000ms

y ~ (time < mu) * (exp(-1 * (time - mu) ^ 2 / (2 * sig1 ^ 2)) * (ht - base1) + base1) + (mu <= time) * (exp(-1 * (time - mu) ^ 2 / (2 * sig2 ^ 2)) * (ht - base2) + base2)

Effective Alpha Functional

Description

Functional that returns function for computing effective alpha for given parameters and distribution

Usage

effectiveAlpha_f(rho, n = 10, df = NULL, method = "norm")

Arguments

Exponential curve function

Description

Exponential function used in fitting nlme curve for observations

Usage

expCurve(dat, y, time, params = NULL, ...)

Arguments

Details

Remove any values of zero, or jitter, before using with bdotsFit

y ~ x_0 exp(k beta)

Find modified alpha

Description

find modified alpha

Usage

findModifiedAlpha(
  rho,
  n,
  df,
  alpha = 0.05,
  errorAcc = 0.001,
  gradDiff = ifelse(cores > 3, 0.5, 0.1),
  cores = 0,
  verbose = FALSE,
  method = "t"
)

Arguments

fwerAlpha

Description

Family wise alpha calculation

Usage

fwerAlpha(rho, k, n = 10)

Arguments

Details

Returns effective alpha, given number of tests and the correlation coefficient. This isn't explicitly checked, but there is no reason this function should take any non-scalar values. Derivation of this can be found on pg 12 of Jake's 'Detecting time-specific differences'. This function performs the expression

1 - P(I_t)P(I_t \ | \ I_{t-1})^{N-1}

Get Fit Correlations

Description

Helper function for finding correlation of fixed value and fitted values within group

Usage

getFitCorforGroups(x, val, ciBands = FALSE, method = "pearson")

Arguments

Return fitted values

Description

Returns fitted values at observed times

Usage

getSubCurveValues(bd, origNames = TRUE, origTime = TRUE)

Arguments

Details

Given a single row of bdObj, this returns fitted values at the observed times to use in conjunction with whatever else

Linear curve function

Description

Linear function used in fitting nlme curve for observations

Usage

linear(dat, y, time, params = NULL, ...)

Arguments

Details

Don't use this function please

y ~ slope*time + intercept

Logistic curve function for nlme

Description

Logistic function used in fitting nlme curve for observations

Usage

logistic(dat, y, time, params = NULL, ...)

Arguments

Details

y ~ mini + (peak - mini) / (1 + exp(4 * slope * (cross - (time)) / (peak - mini)))

Adjust P-values for Multiple Comparisons

Description

Identical to stats::p.adjust, but includes method = "oleson"

Usage

p_adjust(p, method = "oleson", n = length(p), alpha = 0.05, df, rho, cores = 0)

Arguments

Details

This function works identically to the function p.adjust, with the additional option to use method = "oleson". For this option, user must include a value for df, alpha. If method = "oleson" and no value is given for rho, 0.9 will be used. To compute a value for rho from t-statistics, use ar1Solver.

Value

Returns a vector of adjusted p-values just as in p.adjust, but with additional attributes for alphastar and rho.

See Also

ar1Solver

Parameter t-test

Description

Perform t-test on curve parameters of bdotsFit object

Usage

parTest2(bdObj, group, vals = NULL)

Arguments

Details

Performs pairwise t-test. Currently only tests at alpha = 0.95. Also currently only allows t-test within single grouping column. Ability to test across grouping columns to come later

Value

List of t-test results of class bdotsPars_ttest

Examples

## Not run: 
res <- bdotsFit(data = cohort_unrelated,
                subject = "Subject",
                time = "Time",
                y = "Fixations",
                group = c("Group", "LookType"),
                curveType = doubleGauss(concave = TRUE),
                cor = TRUE,
                numRefits = 2,
                cores = 0,
                verbose = FALSE)
tstats <- parTest(res, group = "LookType", vals = c("Cohort", "Unrelated_Cohort"))

## End(Not run)

Plot for object of class bdotsBootObj

Description

Allows a number of different but also unstable option for plotting an object of class bdotsBoot

Usage

## S3 method for class 'bdotsBootObj'
plot(x, alpha = NULL, ciBands = TRUE, plotDiffs = TRUE, group = NULL, ...)

Arguments

Details

This plot function is also a bit unstable and is expected to change

Value

List of ggplot objects, which may be helpful if the margins are weird

Plots for bdotsCorr

Description

Plots correlation of fixed value wtih fitted curves over time

Usage

## S3 method for class 'bdotsCorrObj'
plot(x, ciBands = FALSE, window = NULL, ...)

Arguments

Plot a bdotsFit object

Description

Plot individual fits or model fit parameters from an object of class 'bdotsObj'. These functions are not very stable

Usage

## S3 method for class 'bdotsObj'
plot(x, fitCode = NULL, gridSize = NULL, plotfun = "fits", ...)

Arguments

Details

Right now, these functions are a bit unstable and expected to change. The largest current issue is with the placement of the legend, which cannot be adjusted. If you are running into issues with seeing things correctly, try making the "Plots" window in RStudio larger before running this function

Value

This will return a list of all of the plots rendered.

Polynomial curve function for nlme

Description

Polynomial function used in fitting nlme curve for observations

Usage

polynomial(dat, y, time, degree, raw = TRUE, params = NULL, ...)

Arguments

Details

It's recommended that one uses raw polynomials for this function for numerical stability. As inference is not performed on the parameters themselves, this should have minimial consequences

y ~ mini + (peak - mini) / (1 + exp(4 * slope * (cross - (time)) / (peak - mini)))

Print 'bdotsBootObj'

Description

Prints argument. Really, just the summary function

Usage

## S3 method for class 'bdotsBootObj'
print(x, ...)

Arguments

Details

Generic for printing 'bdotsBootObj'

Print bdotsBoot Summary

Description

That's pretty much it. This is a print method, so there is likely not much need to call it directly

Usage

## S3 method for class 'bdotsBootSummary'
print(x, ...)

Arguments

Print Parameter Test Summary

Description

Print Parameter Test Summary

Usage

## S3 method for class 'bdotsPars_ttest2'
print(x, ...)

Arguments

Details

That's pretty much it. This is a print method, so there is likely not much need to call it directly

Print bdotsObj Summary

Description

Print bdotsObj Summary

Usage

## S3 method for class 'bdotsSummary'
print(x, ...)

Arguments

Details

That's pretty much it. This is a print method, so there is likely not much need to call it directly

rbindlist for bdotsObjects

Description

Similar to data.table::rbindlist, but preserves botsObjects attributes

Usage

## S3 method for class 'bdObjList'
rbindlist(x, ...)

Arguments

Split object of class bdotsObj

Description

Analogous to other splitting functions, but retains necessary attributes across the split object. As of now, it can only be unsplit with bdots::rbindlist

Usage

## S3 method for class 'bdotsObj'
split(x, f, drop = FALSE, by, ...)

Arguments

Subset a nested group bdotsBoot objects

Description

Subset a nested group bdotsBoot objects

Usage

## S3 method for class 'bdotsBootObj'
subset(x, group, adjustAlpha = NULL, ...)

Arguments

Details

This function is used to subset a bdotsBootObject that was fit to compute the difference of differences. This allows the user to subset out the outer group in the comparison for plotting and investigation

Summary for bdotsBootObj

Description

Provides summary information for bdotsBootObj

Usage

## S3 method for class 'bdotsBootObj'
summary(object, ...)

Arguments

Value

Returns an object of class "bdotsBootSummary". There is some summarized information included if assigned to an object, i.e., 'summ <- summary(bdBootObj)' then 'str(summ)'

Summary for bdotsObj

Description

Provides summary information for bdotsObj

Usage

## S3 method for class 'bdotsObj'
summary(object, ...)

Arguments

Value

Returns an object of class "bdotsSummary". There is some summarized information included if assigned to an object, i.e., 'summ <- summary(bdObj)' then 'str(summ)'

target dataset

Description

target dataset - need to include details

Usage

target

Format

An object of class data.frame with 25050 rows and 4 columns.

Write fits from bdotsBoot to csv file

Description

The function is used to write out columns for each group for which a curve was bootstrapped

Usage

writeCSV(bootObj, file, alpha = 0.05, ...)

Arguments

Details

This is potentially useful for constructing plots in a separate application. There is an additional column, Significant indicating if a particular time point was considered significant between the difference curves. For difference of difference objects, this only indicates significance for the outer difference.