About sim_slopes function to conduct clustered robust

[JJery-web]

Hello, help! I'm wondering how to use the sim_slopes function to conduct clustered robust plotting based on the gvkey of enterprises. I've tried using cluster, v.cov, and v.cov.args, but none of them worked:

stata code: reg y c.x##c.mod i.fyear i.industry, cluster(gvkey)

fit1 <- lm(y ~ x * mod + factor(industry) + factor(fyear), data = data)
cov_matrix <- vcovCL(fit1, cluster = data$gvkey)
cov_matrix2 = coeftest(fit1, vcov = cov_matrix)

sim_slopes(fit1, pred = x, modx = mod, jnplot = TRUE, cluster = "gvkey", data = merged_data) 
# There was no error message, but the plotting result was the same as when not adding cluster = "gvkey". I guess it didn't take effect.

sim_slopes(fit1, pred = x, modx = mod, jnplot = TRUE, v.cov = sandwich::vcovCL(fit1, cluster = data$gvkey), v.cov.args = fit1) 
# Error information: vcovargs[[which(sapply(vcovargs, function(x) length(x[[1]]) == : attempt to select less than one element in OneIndex

sim_slopes(fit1, 
           pred = x, 
           modx = mod, 
           jnplot = TRUE, 
           v.cov = sandwich::vcovCL(fit1, cluster = data$gvkey),  # Specify the function for dynamically calculating the covariance
           v.cov.args = fit1)  # Specify the parameters for vcovCL
# Error information: vcovargs[[which(sapply(vcovargs, function(x) length(x[[1]]) == : attempt to select less than one element in OneIndex

[Imane-code-ux]

Hello,

I'm new to this package and, under the supervision of @katossky and @JMeyer31, I’d love to contribute to this issue as part of an assignment at ENSAI (a statistical engineering school in France).

We've gone through the contributing guidelines and will do our best to be independent, but we might reach out for guidance now and then. We have a solid background in statistics and a reasonable one in computer science.

If this issue seems too complex for beginners, please let us know! We'd really appreciate any pointers toward a more suitable issue where we can be helpful.

Looking forward to contributing and learning from the community!

Thanks! 😊
Imane-code-ux

[Imane-code-ux]

So, I was diving into the sim_slopes function in the interactions package and it turns out that sim_slopes handles the vcov argument on its own. This means if you try to pass your own vcov parameter, it might cause some unexpected hiccups.

[Imane-code-ux]

The sim_slopes function is part of the interactions package, which is a spinoff from the jtools package. It is no longer included in jtools. Therefore, when using sim_slopes, ensure that you have the interactions package installed and loaded.​

[Imane-code-ux]

To address this issue, I propose the following steps:

1. Update Documentation: Clearly state in the sim_slopes documentation that the function internally manages the vcov argument and that users should avoid specifying it explicitly. This clarification will help prevent misunderstandings and misuse.​

2. Modify Function Behavior: Implement a check within sim_slopes to detect when a user specifies the vcov argument. If detected, the function should issue a warning or error message informing the user that the vcov parameter is internally handled and should not be provided. This proactive approach will enhance user experience and reduce potential errors.

[Imane-code-ux]

​Since the sim_slopes function is now part of the interactions package, I initially planned to fork that repository to address the issue directly. However, I discovered that you've already added this issue to the interactions package. Therefore, I'll proceed by forking from there and contribute towards resolving the problem.