gvlma() works for panel data and time series?

[Lucas11]

Hi, anybody knows if the gvlma() function to check the assumptions of the linear model works for panel data or for time series?

[technocrat]

It will work with data frame representing chronologically ordered observations but not with time series objects. Mechanically.

library(gvlma)
fit <- lm(Ozone ~ Solar.R + Wind + Temp, data = airquality)
gvlma(fit)
#> 
#> Call:
#> lm(formula = Ozone ~ Solar.R + Wind + Temp, data = airquality)
#> 
#> Coefficients:
#> (Intercept)      Solar.R         Wind         Temp  
#>   -64.34208      0.05982     -3.33359      1.65209  
#> 
#> 
#> ASSESSMENT OF THE LINEAR MODEL ASSUMPTIONS
#> USING THE GLOBAL TEST ON 4 DEGREES-OF-FREEDOM:
#> Level of Significance =  0.05 
#> 
#> Call:
#>  gvlma(x = fit) 
#> 
#>                        Value   p-value                   Decision
#> Global Stat        111.50301 0.000e+00 Assumptions NOT satisfied!
#> Skewness            34.34297 4.621e-09 Assumptions NOT satisfied!
#> Kurtosis            50.45786 1.217e-12 Assumptions NOT satisfied!
#> Link Function       26.64685 2.442e-07 Assumptions NOT satisfied!
#> Heteroscedasticity   0.05533 8.140e-01    Assumptions acceptable.
par(mfrow = c(2,2))
plot(fit)

^{Created on 2023-05-31 with reprex v2.0.2}

This data will seldom pass due to the non-independence of residuals. There are alternatives that take into account the temporal aspect. See the chapter on time series linear regression in the fpp3 text.