Title: Logistic Knowledge Tracing
Version: 1.7.0
Description: Computes Logistic Knowledge Tracing ('LKT') which is a general method for tracking human learning in an educational software system. Please see Pavlik, Eglington, and Harrel-Williams (2021) https://ieeexplore.ieee.org/document/9616435. 'LKT' is a method to compute features of student data that are used as predictors of subsequent performance. 'LKT' allows great flexibility in the choice of predictive components and features computed for these predictive components. The system is built on top of 'LiblineaR', which enables extremely fast solutions compared to base glm() in R.
License: GPL-3
Encoding: UTF-8
LazyData: true
VignetteBuilder: knitr
RoxygenNote: 7.2.3
Depends: R (≥ 3.5.0), SparseM (≥ 1.83), methods, Matrix, data.table (≥ 1.13.2), LiblineaR (≥ 2.10-8)
Imports: glmnet (≥ 4.0-2), glmnetUtils (≥ 1.1.8), lme4 (≥ 1.1-23), cluster (≥ 2.1.3), pROC (≥ 1.16.2), crayon, HDInterval (≥ 0.2.2)
Suggests: rmarkdown, knitr, utils, caret, ggplot2
NeedsCompilation: no
Packaged: 2024-07-01 19:00:46 UTC; ppavl
Author: Philip I. Pavlik Jr. ORCID iD [aut, ctb, cre], Luke G. Eglington ORCID iD [aut, ctb]
Maintainer: Philip I. Pavlik Jr. <imrryr@gmail.com>
Repository: CRAN
Date/Publication: 2024-07-01 22:40:12 UTC

LASSOLKTData

Description

Forward and backwards stepwise search for a set of features and components

with tracking of nonlinear parameters.

Usage

LASSOLKTData(
  data,
  gridpars,
  allcomponents,
  allfeatures,
  preset = NA,
  presetint = T,
  specialcomponents = c(),
  specialfeatures = c(),
  specialpars = c(),
  removefeat = c(),
  removecomp = c()
)

Arguments

data

is a dataset with Anon.Student.Id and CF..ansbin.

gridpars

a vector of parameters to create each feature at

allcomponents

is search space for LKT components

allfeatures

is search space for LKT features

preset

One of "static","AFM","PFA","advanced","AFMLLTM","PFALLTM","advancedLLTM"

presetint

should the intercepts be included for preset components

specialcomponents

add special components (not crossed with features, only paired with special features 1 for 1)

specialfeatures

features for each special component (not crossed during search)

specialpars

parameters for the special features (if needed)

removefeat

Character Vector | Excludes specified features from the test list.

removecomp

Character Vector | Excludes specified components from the test list.

Value

data which is the same frame with the added spacing relevant columns.

list of values "tracetable" and "currentfit"

LASSOLKTModel

Description

runs LASSO search on the data

Usage

LASSOLKTModel(
  data,
  gridpars,
  allcomponents,
  preset = NA,
  presetint = T,
  allfeatures,
  specialcomponents = c(),
  specialfeatures = c(),
  specialpars = c(),
  target_n,
  removefeat = c(),
  removecomp = c(),
  test_fold = 1
)

Arguments

data

is a dataset with Anon.Student.Id and CF..ansbin.

gridpars

a vector of parameters to create each feature at

allcomponents

is search space for LKT components

preset

One of "static","AFM","PFA","advanced","AFMLLTM","PFALLTM","advancedLLTM"

presetint

should the intercepts be included for preset components

allfeatures

is search space for LKT features

specialcomponents

add special components (not crossed with features, only paired with special features 1 for 1)

specialfeatures

features for each special component (not crossed during search)

specialpars

parameters for the special features (if needed)

target_n

chosen number of features in model

removefeat

Character Vector | Excludes specified features from the test list.

removecomp

Character Vector | Excludes specified components from the test list.

test_fold

the fold that the chosen LASSO model will be tested on

Value

list of matrices and values "train_x","train_y","test_x","test_y","fit","target_auc","target_rmse","n_features","auc_lambda","rmse_lambda","BIC_lambda","target_idx", "preds"

LKT

Description

Compute a logistic regression model of learning for input data.

Usage

LKT(
  data,
  usefolds = NA,
  components,
  features,
  fixedpars = NA,
  seedpars = NA,
  interacts = NA,
  curvefeats = NA,
  dualfit = FALSE,
  interc = FALSE,
  verbose = TRUE,
  epsilon = 1e-04,
  cost = 512,
  lowb = 1e-05,
  highb = 0.99999,
  type = 0,
  maketimes = FALSE,
  bias = 0,
  maxitv = 100,
  factrv = 1e+12,
  nosolve = FALSE,
  autoKC = rep(0, length(components)),
  autoKCcont = rep("NA", length(components)),
  connectors = rep("+", max(1, length(components) - 1))
)

Arguments

data

A dataset with Anon.Student.Id and CF..ansbin.

usefolds

Numeric Vector | Specifies the folds for model fitting in LKT; the features are still calculated across all folds to compute test fold fit externally

components

A vector of factors that can be used to compute each features for each subject.

features

a vector methods to use to compute a feature for the component.

fixedpars

a vector of parameters for all features+components.

seedpars

a vector of parameters for all features+components to seed non-linear parameter search.

interacts

A list of components that interacts with component by feature in the main specification.

curvefeats

vector of columns to use with "diff" functions

dualfit

TRUE or FALSE, fit a simple latency using logit. Requires Duration..sec. column in data.

interc

TRUE or FALSE, include a global intercept.

verbose

provides more output in some cases.

epsilon

passed to LiblineaR

cost

passed to LiblineaR

lowb

lower bound for non-linear optimizations

highb

upper bound for non-linear optimizations

type

passed to LiblineaR

maketimes

Boolean indicating whether to create time based features (or may be precomputed)

bias

passed to LiblineaR

maxitv

passed to nonlinear optimization a maxit control

factrv

controls the optim() function

nosolve

causes the function to return a sparse data matrix of the features, rather than a solution

autoKC

a vector to indicate whether to use autoKC for the component (0) or the k for the numebr of clusters

autoKCcont

a vector of text strings set to "rand" for component to make autoKC assignment to cluster is randomized (for comaprison)

connectors

a vector if linear equation R operators including +, * and :

Value

list of values "model", "coefs", "r2", "prediction", "nullmodel", "latencymodel", "optimizedpars","subjectrmse", "newdata", and "automat"

LKT_HDI

Description

Bootstrap credibility intervals to aid in interpreting coefficients.

Usage

LKT_HDI(
  dat,
  n_boot,
  n_students,
  comps,
  feats,
  conns = rep("+", max(1, length(comps) - 1)),
  ints = NA,
  fixeds,
  get_hdi = TRUE,
  cred_mass = 0.95
)

Arguments

dat

Dataframe

n_boot

Number of subsamples to fit

n_students

Number of students per subsample

comps

Components in model

feats

Features in model

conns

R notation for linear equation connectors in model

ints

Interacts in model

fixeds

Fixed parameters in model

get_hdi

Boolean to decide if generating HDI per coefficient

cred_mass

Credibility mass parameter to decide width of HDI

Value

List of values "par_reps", "mod_full", "coef_hdi"

ViewExcel

Description

ViewExcel

Usage

ViewExcel(df = .Last.value, file = tempfile(fileext = ".csv"))

Arguments

df

Dataframe

file

name of the Excel file

buildLKTModel

Description

Forward and backwards stepwise search for a set of features and components

with tracking of nonlinear parameters.

Usage

buildLKTModel(
  data,
  usefolds = NA,
  allcomponents,
  allfeatures,
  currentcomponents = c(),
  specialcomponents = c(),
  specialfeatures = c(),
  forv,
  bacv,
  preset = NA,
  presetint = T,
  currentfeatures = c(),
  verbose = FALSE,
  currentfixedpars = c(),
  maxitv = 10,
  interc = FALSE,
  forward = TRUE,
  backward = TRUE,
  metric = "BIC",
  removefeat = c(),
  removecomp = c()
)

Arguments

data

is a dataset with Anon.Student.Id and CF..ansbin.

usefolds

Numeric Vector | Specifies the folds for model fitting in LKT; the features are still calculated across all folds to compute test fold fit externally

allcomponents

is search space for LKT components

allfeatures

is search space for LKT features

currentcomponents

components to start search from

specialcomponents

add special components (not crossed with features, only paired with special features 1 for 1)

specialfeatures

features for each special component (not crossed during search)

forv

the minimuum amount of improvement needed for the addition of a new term

bacv

the maximuum amount of loss for a term to be removed

preset

One of "static","AFM","PFA","advanced","AFMLLTM","PFALLTM","advancedLLTM"

presetint

should the intercepts be included for preset components

currentfeatures

features to start search from

verbose

passed to LKT

currentfixedpars

used for current features as an option to start

maxitv

passed to LKT

interc

passed to LKT

forward

TRUE or FALSE

backward

TRUE or FALSE

metric

One of "BIC","AUC","AIC", and "RMSE"

removefeat

Character Vector | Excludes specified features from the test list.

removecomp

Character Vector | Excludes specified components from the test list.

Value

list of values "tracetable" and "currentfit"

computeSpacingPredictors

Description

Compute repetition spacing time based features from input data CF..Time. and/or CF..reltime.

which will be automatically computed from Duration..sec. if not present themselves.

Usage

computeSpacingPredictors(data, KCs)

Arguments

data

is a dataset with Anon.Student.Id and CF..ansbin.

KCs

are the components for which spaced features will be specified in LKT

Value

data which is the same frame with the added spacing relevant columns.

computefeatures

Description

Compute feature describing prior practice effect.

Usage

computefeatures(data, feat, par1, par2, index, index2, par3, par4, par5, fcomp)

Arguments

data

copy of main data frame.

feat

is the feature to be computed.

par1

nonlinear parameters used for nonlinear features.

par2

nonlinear parameters used for nonlinear features.

index

a student by component levels index

index2

a component levels index

par3

nonlinear parameters used for nonlinear features.

par4

nonlinear parameters used for nonlinear features.

par5

nonlinear parameters used for nonlinear features.

fcomp

the component name.

Value

a vector suitable for regression input.

countOutcome

Description

Compute the prior sum of the response appearing in the outcome column for the index

Usage

countOutcomeold(data, index, response)

Arguments

data

the dataset to compute an outcome vector for

index

the subsets to count over

response

the actually response value being counted

Value

the vector of the lagged cumulative sum.

Trial sequences for practice participants.

Description

A dataset containing a raw sample from the Memphis Datashop.

Usage

largerawsample

Format

A data frame please see the DataShop for more info.

It has many columns.

Source

https://pslcdatashop.web.cmu.edu/Export?datasetId=5513

Predict for LKT Models

Description

Generates predictions and evaluates logistic regression models tailored for learning data, specifically designed for Logistic Knowledge Tracing (LKT) models. This function provides flexibility in returning either just the predicted probabilities or both the predictions and key evaluation statistics.

Usage

predict_lkt(
  modelob,
  data,
  fold = NULL,
  return_stats = FALSE,
  min_pred_limit = 1e-05,
  max_pred_limit = 0.99999
)

Arguments

modelob

An LKT model object containing necessary model coefficients and predictors for generating predictions.

data

A dataset including predictor variables, the outcome variable CF..ansbin., and fold information.

fold

Optional. Numeric vector specifying which folds to include for prediction. If NULL or empty, uses all data.

return_stats

Logical. If TRUE, returns both predictions and evaluation statistics (Log-Likelihood, AUC, RMSE, R^2). If FALSE, returns only the predictions.

min_pred_limit

Minimum prediction limit. Default is 0.00001.

max_pred_limit

Maximum prediction limit. Default is 0.99999.

Value

If return_stats is FALSE, returns a list containing:

If return_stats is TRUE, returns a list containing:

Trial sequences for practice participants.

Description

A dataset containing a small sample of participants in a memory experiment.

Usage

samplelkt

Format

A data frame with 2074 rows and many variables:

Anon.Student.Id

unique identifier for each student

Duration..sec.

unique identifier for each student

KC..Default.

unique identifier for each student

Outcome

unique identifier for each student

...

Source

https://pslcdatashop.web.cmu.edu/DatasetInfo?datasetId=5508

smallSet

Description

smallSet

Usage

smallSet(data, nSub)

Arguments

data

Dataframe of student data

nSub

Number of students