All software below is free of use, but please cite it if you use it.
Interactive ('Shiny') web applications:
Chi-bar-square difference test
Phi-and-Psi-Plots and Find DeltaT
Standardizing and/or transforming lagged regression estimates
CT meta-analysis on lagged effects
Their citations can be found in the apps themselves.
For each app, you can find example input files here and in the app itself.
Below, you find more information, also regarding accompanying R code (see also the Software tab).
In case you encounter any difficulties or when you have suggestions for improvement, please e-mail me (R.M.Kuiper@uu.nl).
Chi-bar-square difference test (first release)
You can use this interactive web app to calculate the (robust Satorra-Bentler) Chi-bar-square difference test of the RI-CLPM versus the CLPM (first tab) and in general (second tab). The first is discussed in Hamaker, Kuiper, and Grasman (2015) and the second in Stoel et al. (2006). One can use the 'regular' Chi-square values, but also the robust Satorra-Bentler Chi-square values (then an adjusted version of the Chi-bar-square difference test will be calculated).
The Chi-bar-square difference test should be used when, for example, (random intercept) variances are set to zero (i.e., to their boundary value).
Notably, the usual Chi-square difference test is conservative in that case, which means that it will render a higher p-value than the Chi-bar-square difference test will. Hence:
- If the Chi-square difference test is significant, the Chi-bar-square difference test (i.e., the correct test) will render a lower p-value which is then also significant.
- If the Chi-square difference test is not significant, the p-value of the Chi-bar-square difference test will be lower but it is unknown whether it is significant or not. Thus, the Chi-square difference test may incorrectly favour the smaller model (the CLPM when comparing the RI-CLPM with the CLPM). Hence, one should perform the Chi-bar-square difference test.
R package and function ChiBarSq.DiffTest
library(devtools) # Make sure you have Rtools (and a version which is compatible with your R version).
install_github("rebeccakuiper/ChiBarSq.DiffTest")
library(ChiBarSq.DiffTest)
# In case it does not install, check the error messages to see whether some packages need to be installed first.
?ChiBarSq.DiffTest
citation("ChiBarSq.DiffTest")
Notably, in case of problems with installing this 'ChiBarSq.DiffTest' package, it might be needed to install the packages 'ic.infer' and 'nleqslv' yourself (check the error messages in R to see whether and which packages are missing).
Phi-and-Psi-Plots and Find DeltaT (first release)
You can use this interactive web app to obtain more insight into first-order discrete-time model (i.e., VAR(1)) and first-order continuous-time (CT(1)) model parameters. This interactive web app:
More details can be found in: Kuiper, R. M., and Hamaker, E.L. (unpublished). Correlated residuals: What they do (not) represent.
R functions
library(devtools) # Make sure you have Rtools (and a version which is compatible with your R version).
install_github("rebeccakuiper/CTmeta")
# In case it does not install, check the error messages to see whether some packages need to be installed first.
library(CTmeta)
# To obtain more information about the input including examples.
?PhiPlot # to create a Phi-plot
?MaxDeltaT.CTM # to determine for which time-interval Phi(deltaT) reaches minimum/maximum
Here (on my GitHub page), you can also find a (html and pdf) vignette/tutorial for CTmeta ("Introduction to CTmeta...").
lsf.str("package: CTmeta ") # Gives all functions, see below.
citation("CTmeta")
All functions in CTmeta:
?Area
?ChecksCTM
?CTmeta
?CTMparam
?DiagDeltaT
?Gamma.fromCTM
?Gamma.fromVAR
?ggPhiPlot
?MaxDeltaT
?PhiPlot
?SigmaVARPlot
?StandPhi
?StandTransPhi
?TransPhi_Corr
?VARparam
Standardizing and/or transforming lagged regression estimates (first release)
You can use this interactive web app to obtain:
The first is discussed in Kuiper and Hamaker (unpublished), the second in Kuiper and Ryan (2018), and the third in Kuiper and Ryan (2020).
R functions
library(devtools) # Make sure you have Rtools (and a version which is compatible with your R version).
install_github("rebeccakuiper/CTmeta")
# In case it does not install, check the error messages to see whether some packages need to be installed first.
library(CTmeta)
# To obtain more information about the input including examples.
?CovMxStandTransPhi
?CovMxStandPhi
?TransPhi_Corr
A vignette/tutorial with descriptions and examples of the functions can be found here.
lsf.str("package: CTmeta ") # Gives all functions, as denoted above.
citation("CTmeta")
CT meta-analysis on lagged effects (first release)
You can use this interactive web app to meta-analyze standardized lagged effects (based on an underlying continuous-time model). See also the R function 'CTmeta'. It can take the following input:
For more details, see Kuiper and Ryan (2020) and Kuiper (2021).
R package and function CTmeta
library(devtools) # Make sure you have Rtools (and a version which is compatible with your R version).
install_github("rebeccakuiper/CTmeta")
# In case it does not install, check the error messages to see whether some packages need to be installed first.
library(CTmeta)
?CTmeta # To obtain more information about the input including examples.
A vignette/tutorial with descriptions and examples of the functions can be found here.
citation("CTmeta")
ICweights (This is not an interactive web application, only an R package/function)
You can use the R function IC.weights to transform information criteria values (AIC, BIC, GORIC, GORICA) into information criteria weights (IC weights; cf. Akaike weights, AIC weights, GORIC weights, GORICA weights). These weights help in interpretating differences in an information criterion (IC) for multiple models. The IC values themselves can only reflect which model is better (the lower, the better), but cannot quantify by how much. IC weights do: They quantify the number of times one model is better than another one.
You can use the R function ExpectedHypoRate to calculate, among other things, the expected percentage of times hypotheses are chosen under a specific (null) distribution.
R package ICweights and functions IC.weights and ExpectedHypoRate
library(devtools) # Make sure you have Rtools (and a version which is compatible with your R version).
install_github("rebeccakuiper/ICweights")
library(ICweights)
# In case it does not install, check the error messages to see whether some packages need to be installed first.
?IC.weights
?ExpectedHypoRate
citation("ICweights")