R syntax workaround for bug in Custom Contrasts in (classical) Linear Mixed Models?
I am attempting a workaround for a current bug in Classical Linear Mixed Models wherein in the number of allowable custom contrasts appears to be arbitrarily limited.
My attempted workaround uses JASP's generated R syntax to try to circumvent the GUI limitation on the number of contrasts. I already have some custom contrasts implemented in the model (using the GUI), but when I look at the generated syntax, I don't see a specification of those custom contrasts, so I can't figure out what part of the code I need to modify/elaborate in order to produce the entire set of desired custom contrasts. (Is the problem that the generated code shows Contrasts = NULL when it should not be NULL?) The generated syntax is shown below.
jaspMixedModels::MixedModelsLMM(
data = NULL,
version = "0.95.5",
formula = `log RT` ~ Congruence * Tempo * Angle + (1 | Participant),
bootstrapSamples = 1000,
contrasts = NULL,
factorContrast = "sum",
fixedEffectEstimate = FALSE,
includeIntercept = TRUE,
interceptTest = FALSE,
marginalMeansCiLevel = 0.95,
marginalMeansComparison = FALSE,
marginalMeansComparisonWith = 0,
marginalMeansContrast = TRUE,
marginalMeansDf = "satterthwaite",
marginalMeansDfEstimated = FALSE,
marginalMeansPAdjustment = "none",
marginalMeansSd = 1,
marginalMeansTerms = ~ Tempo + Congruence,
modelSummary = TRUE,
plotBackgroundColor = "darkgrey",
plotBackgroundData = ~ Participant.nominal,
plotBackgroundElement = "jitter",
plotCiLevel = 0.95,
plotCiType = "model",
plotDodge = 0.3,
plotElementWidth = 1,
plotEstimatesTable = FALSE,
plotHeight = 320,
plotHorizontalAxis = NULL,
plotJitterHeight = 0,
plotJitterWidth = 0.1,
plotLegendPosition = "none",
plotLevelsByColor = FALSE,
plotLevelsByFill = FALSE,
plotLevelsByLinetype = TRUE,
plotLevelsByShape = TRUE,
plotRelativeSizeData = 1,
plotRelativeSizeText = 1.5,
plotSeparateLines = NULL,
plotSeparatePlots = NULL,
plotTheme = "jasp",
plotTransparency = 0.7,
plotWidth = 480,
randomEffectEstimate = FALSE,
seed = 1,
setSeed = TRUE,
testMethod = "parametricBootstrap",
trendsCiLevel = 0.95,
trendsComparison = FALSE,
trendsComparisonWith = 0,
trendsContrast = FALSE,
trendsContrasts = NULL,
trendsDf = "asymptotic",
trendsDfEstimated = FALSE,
trendsPAdjustment = "holm",
trendsSd = 1,
trendsTrendVariable = NULL,
trendsVariables = NULL,
type = "3",
varianceCorrelationEstimate = FALSE,
vovkSellke = FALSE)
R
Comments
That was a good idea of using the syntax, but I also can't figure out how to make it work. In the meantime here's two options I thought of that you might have thought of already:
option #1 is clunky and inconvenient but much simpler than option #2 it seems to me
Thanks for the suggestions. I complication of Option 1 is that I would need to specify no correction for familywise error, and then semi-manually adjust the p values afterward.
For now, I think I'll just wait and see if the bug gets fixed soon.
R
ah yes you're right didn't think about that. OK, if you change your mind later and want to try the R console I should be able to help