Kliegl et al. (2015) is a follow-up to Kliegl et al. (2011) (see also script kwdyz11.qmd) from an experiment looking at a variety of effects of visual cueing under four different cue-target relations (CTRs). In this experiment two rectangles are displayed (1) in horizontal orientation , (2) in vertical orientation, (3) in left diagonal orientation, or in (4) right diagonal orientation relative to a central fixation point. Subjects react to the onset of a small or a large visual target occurring at one of the four ends of the two rectangles. The target is cued validly on 70% of trials by a brief flash of the corner of the rectangle at which it appears; it is cued invalidly at the three other locations 10% of the trials each. This implies a latent imbalance in design that is not visible in the repeated-measures ANOVA, but we will show its effect in the random-effect structure and conditional modes.
There are a couple of differences between the first and this follow-up experiment, rendering it more a conceptual than a direct replication. First, the original experiment was carried out at Peking University and this follow-up at Potsdam University. Second, diagonal orientations of rectangles and large target sizes were not part of the design of Kliegl et al. (2011).
We specify three contrasts for the four-level factor CTR that are derived from spatial, object-based, and attractor-like features of attention. They map onto sequential differences between appropriately ordered factor levels. Replicating Kliegl et al. (2011), the attraction effect was not significant as a fixed effect, but yielded a highly reliable variance component (VC; i.e., reliable individual differences in positive and negative attraction effects cancel the fixed effect). Moreover, these individual differences in the attraction effect were negatively correlated with those in the spatial effect.
This comparison is of interest because a few years after the publication of Kliegl et al. (2011), the theoretically critical correlation parameter (CP) between the spatial effect and the attraction effect was determined as the source of a non-singular LMM in that paper. The present study served the purpose to estimate this parameter with a larger sample and a wider variety of experimental conditions.
Here we also include two additional experimental manipulations of target size and orientation of cue rectangle. A similar analysis was reported in the parsimonious mixed-model paper (Bates et al., 2015); it was also used in a paper of GAMEMs (Baayen et al., 2017). Data and R scripts of those analyses are also available in R-package RePsychLing.
The analysis is based on reaction times rt to maintain compatibility with Kliegl et al. (2011).
In this vignette we focus on the reduction of model complexity. And we start with a quote:
“Neither the [maximal] nor the [minimal] linear mixed models are appropriate for most repeated measures analysis. Using the [maximal] model is generally wasteful and costly in terms of statiscal power for testing hypotheses. On the other hand, the [minimal] model fails to account for nontrivial correlation among repeated measurements. This results in inflated [T]ype I error rates when non-negligible correlation does in fact exist. We can usually find middle ground, a covariance model that adequately accounts for correlation but is more parsimonious than the [maximal] model. Doing so allows us full control over [T]ype I error rates without needlessly sacrificing power.”
Stroup, W. W. (2012, p. 185). Generalized linear mixed models: Modern concepts, methods and applica?ons. CRC Press, Boca Raton.
using AlgebraOfGraphics
using AlgebraOfGraphics: density
using BoxCox
using CairoMakie
using CategoricalArrays
using Chain
using DataFrameMacros
using DataFrames
using MixedModels
using MixedModelsMakie
using Random
using SMLP2024: dataset
using StatsBasedat = DataFrame(dataset(:kkl15))
describe(dat)| Row | variable | mean | min | median | max | nmissing | eltype |
|---|---|---|---|---|---|---|---|
| Symbol | Union… | Any | Union… | Any | Int64 | DataType | |
| 1 | Subj | S001 | S147 | 0 | String | ||
| 2 | CTR | dod | val | 0 | String | ||
| 3 | rt | 293.147 | 150.22 | 276.594 | 749.481 | 0 | Float32 |
| 4 | cardinal | cardinal | diagonal | 0 | String | ||
| 5 | size | big | small | 0 | String |
dat_subj = combine(
groupby(dat, [:Subj, :CTR]),
nrow => :n,
:rt => mean => :rt_m,
:rt => (c -> mean(log, c)) => :lrt_m,
)
dat_subj.CTR = categorical(dat_subj.CTR, levels=levels(dat.CTR))
describe(dat_subj)| Row | variable | mean | min | median | max | nmissing | eltype |
|---|---|---|---|---|---|---|---|
| Symbol | Union… | Any | Union… | Any | Int64 | DataType | |
| 1 | Subj | S001 | S147 | 0 | String | ||
| 2 | CTR | val | dod | 0 | CategoricalValue{String, UInt32} | ||
| 3 | n | 156.294 | 49 | 64.0 | 448 | 0 | Int64 |
| 4 | rt_m | 308.223 | 208.194 | 304.862 | 584.71 | 0 | Float32 |
| 5 | lrt_m | 5.6908 | 5.33226 | 5.69848 | 6.36141 | 0 | Float32 |
boxplot(
dat_subj.CTR.refs,
dat_subj.lrt_m;
orientation=:horizontal,
show_notch=true,
axis=(;
yticks=(
1:4,
[
"valid cue",
"same obj/diff pos",
"diff obj/same pos",
"diff obj/diff pos",
]
)
),
figure=(; resolution=(800, 300)),
)┌ Warning: Found `resolution` in the theme when creating a `Scene`. The `resolution` keyword for `Scene`s and `Figure`s has been deprecated. Use `Figure(; size = ...` or `Scene(; size = ...)` instead, which better reflects that this is a unitless size and not a pixel resolution. The key could also come from `set_theme!` calls or related theming functions.
└ @ Makie ~/.julia/packages/Makie/WgbrE/src/scenes.jl:227
Mean of reaction times for four cue-target relations. Targets appeared at (a) the cued position (valid) in a rectangle, (b) in the same rectangle cue, but at its other end, (c) on the second rectangle, but at a corresponding horizontal/vertical physical distance, or (d) at the other end of the second rectangle, that is \(\sqrt{2}\) of horizontal/vertical distance diagonally across from the cue, that is also at larger physical distance compared to (c).
contrasts = Dict(
:CTR => SeqDiffCoding(; levels=["val", "sod", "dos", "dod"]),
:cardinal => EffectsCoding(; levels=["cardinal", "diagonal"]),
:size => EffectsCoding(; levels=["big", "small"])
)Dict{Symbol, StatsModels.AbstractContrasts} with 3 entries:
:CTR => SeqDiffCoding(["val", "sod", "dos", "dod"])
:size => EffectsCoding(nothing, ["big", "small"])
:cardinal => EffectsCoding(nothing, ["cardinal", "diagonal"])This is the maximum LMM for the design; size is a between-subject factor, ignoring other information such as trial number, age and gender of subjects.
m_max = let
form = @formula rt ~ 1 + CTR * cardinal * size +
(1 + CTR * cardinal | Subj)
fit(MixedModel, form, dat; contrasts)
end;Minimizing 142 Time: 0:00:00 ( 0.70 ms/it)
objective: 599458.051337806
Minimizing 876 Time: 0:00:00 ( 0.42 ms/it)
Minimizing 877 Time: 0:00:00 ( 0.44 ms/it)
objective: 599358.9785788914issingular(m_max)trueonly(MixedModels.PCA(m_max))
Principal components based on correlation matrix
(Intercept) 1.0 … . . . . .
CTR: sod 0.6 . . . . .
CTR: dos 0.44 . . . . .
CTR: dod 0.38 1.0 . . . .
cardinal: diagonal -0.03 0.15 1.0 . . .
CTR: sod & cardinal: diagonal 0.32 … 0.02 0.26 1.0 . .
CTR: dos & cardinal: diagonal 0.42 -0.02 0.24 0.43 1.0 .
CTR: dod & cardinal: diagonal 0.14 0.88 0.11 0.26 0.07 1.0
Normalized cumulative variances:
[0.3611, 0.5905, 0.7428, 0.8638, 0.9457, 1.0, 1.0, 1.0]
Component loadings
PC1 … PC4 PC5 PC6 PC7 PC8
(Intercept) -0.43 -0.15 -0.29 0.47 -0.19 0.46
CTR: sod -0.35 -0.41 -0.03 -0.23 -0.12 -0.59
CTR: dos -0.39 0.35 -0.1 -0.75 -0.08 0.19
CTR: dod -0.45 0.22 -0.18 0.1 0.7 0.06
cardinal: diagonal -0.12 … -0.56 -0.48 -0.16 -0.13 0.1
CTR: sod & cardinal: diagonal -0.31 -0.43 0.72 -0.01 0.26 0.14
CTR: dos & cardinal: diagonal -0.24 0.22 -0.15 0.33 0.06 -0.6
CTR: dod & cardinal: diagonal -0.41 0.3 0.31 0.17 -0.6 0.0VarCorr(m_max)| Column | Variance | Std.Dev | Corr. | |||||||
| Subj | (Intercept) | 2938.44786 | 54.20745 | |||||||
| CTR: sod | 455.44655 | 21.34119 | +0.60 | |||||||
| CTR: dos | 56.85753 | 7.54039 | +0.44 | +0.18 | ||||||
| CTR: dod | 192.63470 | 13.87929 | +0.38 | +0.54 | +0.30 | |||||
| cardinal: diagonal | 261.91287 | 16.18372 | -0.03 | -0.01 | +0.00 | +0.15 | ||||
| CTR: sod & cardinal: diagonal | 30.20533 | 5.49594 | +0.32 | +0.21 | +0.33 | +0.02 | +0.26 | |||
| CTR: dos & cardinal: diagonal | 1.64655 | 1.28318 | +0.42 | -0.37 | +0.58 | -0.02 | +0.24 | +0.43 | ||
| CTR: dod & cardinal: diagonal | 13.89368 | 3.72742 | +0.14 | +0.28 | +0.25 | +0.88 | +0.11 | +0.26 | +0.07 | |
| Residual | 3972.11165 | 63.02469 |
The LMM m_max is overparameterized but it is not immediately apparent why.
Force CPs to zero. Reduction strategy 1 is more suited for reducing model w/o theoretical expectations about CPs. The better reduction strategy for the present experiment with an a priori interest in CPs is described as Reduction strategy 2.
m_zcp1 = let
form = @formula rt ~ 1 + CTR * cardinal * size +
zerocorr(1 + CTR * cardinal | Subj)
fit(MixedModel, form, dat; contrasts)
end;issingular(m_zcp1)trueonly(MixedModels.PCA(m_zcp1))
Principal components based on correlation matrix
(Intercept) 1.0 . . . . . . .
CTR: sod 0.0 1.0 . . . . . .
CTR: dos 0.0 0.0 1.0 . . . . .
CTR: dod 0.0 0.0 0.0 1.0 . . . .
cardinal: diagonal 0.0 0.0 0.0 0.0 1.0 . . .
CTR: sod & cardinal: diagonal 0.0 0.0 0.0 0.0 0.0 1.0 . .
CTR: dos & cardinal: diagonal 0.0 0.0 0.0 0.0 0.0 0.0 0.0 .
CTR: dod & cardinal: diagonal 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
Normalized cumulative variances:
[0.1429, 0.2857, 0.4286, 0.5714, 0.7143, 0.8571, 1.0, 1.0]
Component loadings
PC1 PC2 … PC4 PC5 PC6 PC7 PC8
(Intercept) 1.0 0.0 0.0 0.0 0.0 0.0 0.0
CTR: sod 0.0 1.0 0.0 0.0 0.0 0.0 0.0
CTR: dos 0.0 0.0 0.0 0.0 0.0 0.0 0.0
CTR: dod 0.0 0.0 1.0 0.0 0.0 0.0 0.0
cardinal: diagonal 0.0 0.0 … 0.0 1.0 0.0 0.0 0.0
CTR: sod & cardinal: diagonal 0.0 0.0 0.0 0.0 1.0 0.0 0.0
CTR: dos & cardinal: diagonal 0.0 0.0 0.0 0.0 0.0 0.0 NaN
CTR: dod & cardinal: diagonal 0.0 0.0 0.0 0.0 0.0 1.0 0.0VarCorr(m_zcp1)| Column | Variance | Std.Dev | Corr. | |||||||
| Subj | (Intercept) | 2875.43244 | 53.62306 | |||||||
| CTR: sod | 483.09089 | 21.97933 | . | |||||||
| CTR: dos | 79.93243 | 8.94049 | . | . | ||||||
| CTR: dod | 216.91700 | 14.72810 | . | . | . | |||||
| cardinal: diagonal | 250.32171 | 15.82156 | . | . | . | . | ||||
| CTR: sod & cardinal: diagonal | 35.91784 | 5.99315 | . | . | . | . | . | |||
| CTR: dos & cardinal: diagonal | 0.00000 | 0.00000 | . | . | . | . | . | . | ||
| CTR: dod & cardinal: diagonal | 6.88389 | 2.62372 | . | . | . | . | . | . | . | |
| Residual | 3972.38114 | 63.02683 |
The LMM m_zcp1 is also overparameterized, but now there is clear evidence for absence of evidence for the VC of one of the interactions and the other two interaction-based VCs are also very small.
Take out VCs for interactions.
m_zcp1_rdc = let
form = @formula rt ~ 1 + CTR * cardinal * size +
zerocorr(1 + CTR + cardinal | Subj)
fit(MixedModel, form, dat; contrasts)
end;issingular(m_zcp1_rdc)falseonly(MixedModels.PCA(m_zcp1_rdc))
Principal components based on correlation matrix
(Intercept) 1.0 . . . .
CTR: sod 0.0 1.0 . . .
CTR: dos 0.0 0.0 1.0 . .
CTR: dod 0.0 0.0 0.0 1.0 .
cardinal: diagonal 0.0 0.0 0.0 0.0 1.0
Normalized cumulative variances:
[0.2, 0.4, 0.6, 0.8, 1.0]
Component loadings
PC1 PC2 PC3 PC4 PC5
(Intercept) 0.0 0.0 0.0 0.0 1.0
CTR: sod 1.0 0.0 0.0 0.0 0.0
CTR: dos 0.0 1.0 0.0 0.0 0.0
CTR: dod 0.0 0.0 1.0 0.0 0.0
cardinal: diagonal 0.0 0.0 0.0 1.0 0.0VarCorr(m_zcp1_rdc)| Column | Variance | Std.Dev | Corr. | ||||
| Subj | (Intercept) | 2881.45052 | 53.67914 | ||||
| CTR: sod | 484.31679 | 22.00720 | . | ||||
| CTR: dos | 79.54961 | 8.91906 | . | . | |||
| CTR: dod | 216.59820 | 14.71728 | . | . | . | ||
| cardinal: diagonal | 244.93991 | 15.65056 | . | . | . | . | |
| Residual | 3980.87777 | 63.09420 |
LMM m_zcp_rdc is ok . We add in CPs.
Extend zcp-reduced LMM with CPs
m_prm1 = let
form = @formula rt ~ 1 + CTR * cardinal * size +
(1 + CTR + cardinal | Subj)
fit(MixedModel, form, dat; contrasts)
end;issingular(m_prm1)falseonly(MixedModels.PCA(m_prm1))
Principal components based on correlation matrix
(Intercept) 1.0 . . . .
CTR: sod 0.6 1.0 . . .
CTR: dos 0.45 0.19 1.0 . .
CTR: dod 0.37 0.54 0.29 1.0 .
cardinal: diagonal -0.1 -0.05 -0.09 0.13 1.0
Normalized cumulative variances:
[0.4488, 0.6668, 0.8307, 0.9437, 1.0]
Component loadings
PC1 PC2 PC3 PC4 PC5
(Intercept) -0.55 -0.16 0.01 0.58 -0.58
CTR: sod -0.54 0.09 -0.49 0.17 0.66
CTR: dos -0.4 -0.25 0.8 -0.13 0.34
CTR: dod -0.49 0.36 -0.1 -0.71 -0.35
cardinal: diagonal 0.05 0.88 0.32 0.34 0.07LMM m_zcp_rdc is ok . We add in CPs.
VarCorr(m_prm1)| Column | Variance | Std.Dev | Corr. | ||||
| Subj | (Intercept) | 2926.39374 | 54.09615 | ||||
| CTR: sod | 454.10921 | 21.30984 | +0.60 | ||||
| CTR: dos | 56.75233 | 7.53341 | +0.45 | +0.19 | |||
| CTR: dod | 188.74621 | 13.73849 | +0.37 | +0.54 | +0.29 | ||
| cardinal: diagonal | 245.27143 | 15.66114 | -0.10 | -0.05 | -0.09 | +0.13 | |
| Residual | 3981.71306 | 63.10082 |
We note that the critical correlation parameter between spatial (sod) and attraction (dod) is now estimated at .54 – not that close to the 1.0 boundary that caused singularity in Kliegl et al. (2011).
gof_summary = let
nms = [:m_zcp1_rdc, :m_prm1, :m_max]
mods = eval.(nms)
lrt = MixedModels.likelihoodratiotest(m_zcp1_rdc, m_prm1, m_max)
DataFrame(;
name = nms,
dof=dof.(mods),
deviance=round.(deviance.(mods), digits=0),
AIC=round.(aic.(mods),digits=0),
AICc=round.(aicc.(mods),digits=0),
BIC=round.(bic.(mods),digits=0),
χ²=vcat(:., round.(lrt.tests.deviancediff, digits=0)),
χ²_dof=vcat(:., round.(lrt.tests.dofdiff, digits=0)),
pvalue=vcat(:., round.(lrt.tests.pvalues, digits=3))
)
end| Row | name | dof | deviance | AIC | AICc | BIC | χ² | χ²_dof | pvalue |
|---|---|---|---|---|---|---|---|---|---|
| Symbol | Int64 | Float64 | Float64 | Float64 | Float64 | Any | Any | Any | |
| 1 | m_zcp1_rdc | 22 | 599486.0 | 599530.0 | 599530.0 | 599725.0 | . | . | . |
| 2 | m_prm1 | 32 | 599418.0 | 599482.0 | 599482.0 | 599766.0 | 68.0 | 10.0 | 0.0 |
| 3 | m_max | 53 | 599359.0 | 599465.0 | 599465.0 | 599936.0 | 59.0 | 21.0 | 0.0 |
AIC prefers LMM m_prm1 over m_zcp1_rdc; BIC LMM m_zcp1_rdc. As the CPs were one reason for conducting this experiment, AIC is the criterion of choice.
Relative to LMM m_max, first we take out interaction VCs and associated CPs, because these VCs are very small. This is the same as LMM m_prm1 above.
m_cpx = let
form = @formula rt ~ 1 + CTR * cardinal * size +
(1 + CTR + cardinal | Subj)
fit(MixedModel, form, dat; contrasts)
end;Now we check the significance of ensemble of CPs.
m_zcp2 = let
form = @formula rt ~ 1 + CTR * cardinal * size +
zerocorr(1 + CTR + cardinal | Subj)
fit(MixedModel, form, dat; contrasts)
end;VarCorr(m_zcp2)| Column | Variance | Std.Dev | Corr. | ||||
| Subj | (Intercept) | 2881.45052 | 53.67914 | ||||
| CTR: sod | 484.31679 | 22.00720 | . | ||||
| CTR: dos | 79.54961 | 8.91906 | . | . | |||
| CTR: dod | 216.59820 | 14.71728 | . | . | . | ||
| cardinal: diagonal | 244.93991 | 15.65056 | . | . | . | . | |
| Residual | 3980.87777 | 63.09420 |
The cardinal-related CPs are quite small. Do we need them?
m_prm2 = let
form = @formula(rt ~ 1 + CTR * cardinal * size +
(1 + CTR | Subj) + (0 + cardinal | Subj))
fit(MixedModel, form, dat; contrasts)
end;VarCorr(m_prm2)| Column | Variance | Std.Dev | Corr. | ||||
| Subj | (Intercept) | 2923.32186 | 54.06775 | ||||
| CTR: sod | 454.38842 | 21.31639 | +0.60 | ||||
| CTR: dos | 56.46065 | 7.51403 | +0.45 | +0.19 | |||
| CTR: dod | 187.89792 | 13.70759 | +0.37 | +0.54 | +0.30 | ||
| cardinal: diagonal | 245.16943 | 15.65789 | . | . | . | . | |
| Residual | 3981.74285 | 63.10105 |
gof_summary = let
nms = [:m_zcp2, :m_prm2, :m_cpx, :m_max]
mods = eval.(nms)
lrt = MixedModels.likelihoodratiotest(m_zcp2, m_prm2, m_cpx, m_max)
DataFrame(;
name = nms,
dof=dof.(mods),
deviance=round.(deviance.(mods), digits=0),
AIC=round.(aic.(mods),digits=0),
AICc=round.(aicc.(mods),digits=0),
BIC=round.(bic.(mods),digits=0),
χ²=vcat(:., round.(lrt.tests.deviancediff, digits=0)),
χ²_dof=vcat(:., round.(lrt.tests.dofdiff, digits=0)),
pvalue=vcat(:., round.(lrt.tests.pvalues, digits=3))
)
end| Row | name | dof | deviance | AIC | AICc | BIC | χ² | χ²_dof | pvalue |
|---|---|---|---|---|---|---|---|---|---|
| Symbol | Int64 | Float64 | Float64 | Float64 | Float64 | Any | Any | Any | |
| 1 | m_zcp2 | 22 | 599486.0 | 599530.0 | 599530.0 | 599725.0 | . | . | . |
| 2 | m_prm2 | 28 | 599420.0 | 599476.0 | 599476.0 | 599725.0 | 65.0 | 6.0 | 0.0 |
| 3 | m_cpx | 32 | 599418.0 | 599482.0 | 599482.0 | 599766.0 | 2.0 | 4.0 | 0.652 |
| 4 | m_max | 53 | 599359.0 | 599465.0 | 599465.0 | 599936.0 | 59.0 | 21.0 | 0.0 |
The cardinal-related CPs could be removed w/o loss of goodness of fit. However, there is no harm in keeping them in the LMM. The data support both LMM m_prm2 and m_cpx (same as: m_prm1). We keep the slightly more complex LMM m_cpx (m_prm1).
Do model residuals meet LMM assumptions? Classic plots are
The slant in residuals show a lower and upper boundary of reaction times, that is we have have too few short and too few long residuals. Not ideal, but at least width of the residual band looks similar across the fitted values, that is there is no evidence for heteroskedasticity.
CairoMakie.activate!(; type="png")
scatter(fitted(m_prm1), residuals(m_prm1); alpha=0.3)
With many observations the scatterplot is not that informative. Contour plots or heatmaps may be an alternative.
set_aog_theme!()
draw(
data((; f=fitted(m_prm1), r=residuals(m_prm1))) *
mapping(
:f => "Fitted values from m1", :r => "Residuals from m1"
) *
density();
)
The plot of quantiles of model residuals over corresponding quantiles of the normal distribution should yield a straight line along the main diagonal.
CairoMakie.activate!(; type="png")
qqnorm(
residuals(m_prm1);
qqline=:none,
axis=(;
xlabel="Standard normal quantiles",
ylabel="Quantiles of the residuals from model m1",
),
)
cm1 = only(ranefinfo(m_prm1))
caterpillar!(Figure(; resolution=(800, 1200)), cm1; orderby=2)┌ Warning: Found `resolution` in the theme when creating a `Scene`. The `resolution` keyword for `Scene`s and `Figure`s has been deprecated. Use `Figure(; size = ...` or `Scene(; size = ...)` instead, which better reflects that this is a unitless size and not a pixel resolution. The key could also come from `set_theme!` calls or related theming functions.
└ @ Makie ~/.julia/packages/Makie/WgbrE/src/scenes.jl:227
shrinkageplot!(Figure(; resolution=(1000, 1200)), m_prm1)┌ Warning: Found `resolution` in the theme when creating a `Scene`. The `resolution` keyword for `Scene`s and `Figure`s has been deprecated. Use `Figure(; size = ...` or `Scene(; size = ...)` instead, which better reflects that this is a unitless size and not a pixel resolution. The key could also come from `set_theme!` calls or related theming functions.
└ @ Makie ~/.julia/packages/Makie/WgbrE/src/scenes.jl:227
Here we
We generate 2500 samples for the 15 model parameters (4 fixed effect, 7 VCs, 15 CPs, and 1 residual).
samp = parametricbootstrap(MersenneTwister(1234321), 2500, m_prm1;
optsum_overrides=(; ftol_rel=1e-8));tbl = samp.tblTable with 48 columns and 2500 rows:
obj β01 β02 β03 β04 β05 β06 ⋯
┌────────────────────────────────────────────────────────────────────
1 │ 5.98897e5 301.101 22.1122 11.2895 0.876659 5.29574 20.7015 ⋯
2 │ 5.9964e5 314.198 23.9774 13.0657 3.54125 6.27177 26.0213 ⋯
3 │ 5.99601e5 313.482 25.8522 12.6864 2.6665 5.99473 26.2079 ⋯
4 │ 5.99422e5 307.342 21.8929 14.5687 5.14975 6.46233 21.4816 ⋯
5 │ 5.9891e5 303.107 22.0525 13.8717 1.22361 9.06571 27.3694 ⋯
6 │ 5.99241e5 304.327 20.6646 12.9426 3.01194 5.19863 31.4934 ⋯
7 │ 599535.0 315.078 26.3107 13.0825 6.77259 6.14113 18.2759 ⋯
8 │ 5.99182e5 306.15 23.4631 12.4967 4.0268 6.32004 27.5211 ⋯
9 │ 6.0014e5 304.214 22.5889 10.4741 3.78498 6.67903 23.2066 ⋯
10 │ 5.994e5 307.004 25.5547 14.8672 1.09287 3.08337 18.9497 ⋯
11 │ 5.98984e5 302.778 20.5811 12.5438 3.28193 7.46265 28.1793 ⋯
12 │ 5.99649e5 313.379 26.6914 10.9561 1.26339 4.65612 32.3437 ⋯
13 │ 5.99031e5 304.034 22.6971 12.175 2.1841 7.84183 36.4001 ⋯
14 │ 5.9947e5 309.43 22.6176 15.1642 -0.58606 3.62731 32.418 ⋯
15 │ 5.99092e5 319.095 24.0423 14.1444 5.27404 8.75307 29.1921 ⋯
16 │ 5.99633e5 310.267 22.907 15.5719 3.72037 8.71994 25.3559 ⋯
17 │ 5.99831e5 307.412 21.8683 13.7834 2.23826 9.10737 30.2628 ⋯
⋮ │ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋱confint(samp)DictTable with 2 columns and 32 rows:
par lower upper
────┬────────────────────
β01 │ 296.443 318.781
β02 │ 18.3734 27.7957
β03 │ 10.3257 16.0712
β04 │ -1.18121 6.18787
β05 │ 3.44619 10.2913
β06 │ 14.9788 37.9644
β07 │ 1.80928 5.35864
β08 │ -0.998456 3.73246
β09 │ -2.53041 2.20858
β10 │ 3.72405 13.3908
β11 │ -3.54059 2.37317
β12 │ 3.72117 11.1719
β13 │ -1.5557 5.11338
β14 │ -2.46016 1.04481
β15 │ -2.68508 2.1626
β16 │ 1.70393 6.52866
ρ01 │ 0.336762 0.792209
⋮ │ ⋮ ⋮We can also visualize the shortest coverage intervals for fixed effects with the ridgeplot() command:
ridgeplot(samp; show_intercept=false)
draw(
data(tbl) *
mapping(:σ => "Residual") *
density();
figure=(; resolution=(800, 400)),
)┌ Warning: Found `resolution` in the theme when creating a `Scene`. The `resolution` keyword for `Scene`s and `Figure`s has been deprecated. Use `Figure(; size = ...` or `Scene(; size = ...)` instead, which better reflects that this is a unitless size and not a pixel resolution. The key could also come from `set_theme!` calls or related theming functions.
└ @ Makie ~/.julia/packages/Makie/WgbrE/src/scenes.jl:227
The shortest coverage interval for the GM ranges from x to x ms and the associate variance component from .x to .x. To keep the plot range small we do not include their densities here.
rn = renamer([
"(Intercept)" => "GM",
"CTR: sod" => "spatial effect",
"CTR: dos" => "object effect",
"CTR: dod" => "attraction effect",
"(Intercept), CTR: sod" => "GM, spatial",
"(Intercept), CTR: dos" => "GM, object",
"CTR: sod, CTR: dos" => "spatial, object",
"(Intercept), CTR: dod" => "GM, attraction",
"CTR: sod, CTR: dod" => "spatial, attraction",
"CTR: dos, CTR: dod" => "object, attraction",
])
draw(
data(tbl) *
mapping(
[:β02, :β03, :β04] .=> "Experimental effect size [ms]";
color=dims(1) =>
renamer(["spatial effect", "object effect", "attraction effect"]) =>
"Experimental effects",
) *
density();
figure=(; resolution=(800, 350)),
)┌ Warning: Found `resolution` in the theme when creating a `Scene`. The `resolution` keyword for `Scene`s and `Figure`s has been deprecated. Use `Figure(; size = ...` or `Scene(; size = ...)` instead, which better reflects that this is a unitless size and not a pixel resolution. The key could also come from `set_theme!` calls or related theming functions.
└ @ Makie ~/.julia/packages/Makie/WgbrE/src/scenes.jl:227
The densitiies correspond nicely with the shortest coverage intervals.
draw(
data(tbl) *
mapping(
[:σ2, :σ3, :σ4] .=> "Standard deviations [ms]";
color=dims(1) =>
renamer(["spatial effect", "object effect", "attraction effect"]) =>
"Variance components",
) *
density();
figure=(; resolution=(800, 350)),
)┌ Warning: Found `resolution` in the theme when creating a `Scene`. The `resolution` keyword for `Scene`s and `Figure`s has been deprecated. Use `Figure(; size = ...` or `Scene(; size = ...)` instead, which better reflects that this is a unitless size and not a pixel resolution. The key could also come from `set_theme!` calls or related theming functions.
└ @ Makie ~/.julia/packages/Makie/WgbrE/src/scenes.jl:227
The VC are all very nicely defined.
draw(
data(tbl) *
mapping(
[:ρ01, :ρ02, :ρ03, :ρ04, :ρ05, :ρ06] .=> "Correlation";
color=dims(1) =>
renamer(["GM, spatial", "GM, object", "spatial, object",
"GM, attraction", "spatial, attraction", "object, attraction"]) =>
"Correlation parameters",
) *
density();
figure=(; resolution=(800, 350)),
)┌ Warning: Found `resolution` in the theme when creating a `Scene`. The `resolution` keyword for `Scene`s and `Figure`s has been deprecated. Use `Figure(; size = ...` or `Scene(; size = ...)` instead, which better reflects that this is a unitless size and not a pixel resolution. The key could also come from `set_theme!` calls or related theming functions.
└ @ Makie ~/.julia/packages/Makie/WgbrE/src/scenes.jl:227
Three CPs stand out positively, the correlation between GM and the spatial effect, GM and attraction effect, and the correlation between spatial and attraction effects. The second CP was positive, but not significant in the first study. The third CP replicates a CP that was judged questionable in script kwdyz11.jl. The three remaining CPs are not well defined for reaction times.