`digits.p = 1` gives incorrect p value
felixgolcher opened this issue · comments
Thanks a lot for that fantastic package. It saved so much time, effort and brain for me!
Setting digits.p to 1 shows (here, at least) pvals to be below .001 which are clearly not:
set.seed(20)
m0 <- lm(extra~group, sleep[sample(nrow(sleep), 20, replace = T),]) # get right amount of data
coefficients(summary(m0))["group2","Pr(>|t|)"] # pval ~ .02
sjPlot::tab_model(m0, digits.p = 1) # shows pval as < .001 ❌
sjPlot::tab_model(m0, digits.p = 2) # shows pval ~ 02 ✅
Session Info:
> sessionInfo()
R version 4.1.1 (2021-08-10)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Manjaro Linux
Matrix products: default
BLAS: /usr/lib/libopenblasp-r0.3.18.so
LAPACK: /usr/lib/liblapack.so.3.10.0
locale:
[1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8 LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
[7] LC_PAPER=en_US.UTF-8 LC_NAME=C LC_ADDRESS=C LC_TELEPHONE=C LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] sjPlot_2.8.9
loaded via a namespace (and not attached):
[1] zoo_1.8-9 tidyselect_1.1.1 sjlabelled_1.1.8 xfun_0.26 performance_0.8.0 purrr_0.3.4 splines_4.1.1 lattice_0.20-45 parameters_0.14.0
[10] colorspace_2.0-2 vctrs_0.3.8 generics_0.1.0 utf8_1.2.2 survival_3.2-13 rlang_0.4.11 nloptr_1.2.2.2 pillar_1.6.3 glue_1.4.2
[19] DBI_1.1.1 modelr_0.1.8 effectsize_0.5 emmeans_1.7.0 multcomp_1.4-17 lifecycle_1.0.1 sjmisc_2.8.7 munsell_0.5.0 gtable_0.3.0
[28] bayestestR_0.11.0 mvtnorm_1.1-3 codetools_0.2-18 coda_0.19-4 knitr_1.36 datawizard_0.2.1 fansi_0.5.0 Rcpp_1.0.7 broom_0.7.9
[37] TH.data_1.1-0 xtable_1.8-4 scales_1.1.1 backports_1.2.1 ggeffects_1.1.1 lme4_1.1-27.1 ggplot2_3.3.5 insight_0.14.4 dplyr_1.0.7
[46] grid_4.1.1 tools_4.1.1 sjstats_0.18.1 sandwich_3.0-1 magrittr_2.0.1 tibble_3.1.5 crayon_1.4.1 tidyr_1.1.4 pkgconfig_2.0.3
[55] MASS_7.3-54 ellipsis_0.3.2 Matrix_1.3-4 estimability_1.3 minqa_1.2.4 assertthat_0.2.1 R6_2.5.1 boot_1.3-28 nlme_3.1-153
[64] compiler_4.1.1
Another, very small point: p-vals are not zero padded, right? With digits.p set to two you get .065, but also .01 instead of the .010 some might prefer.
Setting digits.p to 1 shows (here, at least) pvals to be below .001 which are clearly not:
This function simply works with round(), that's why you get 0, and 0 is formatted a <0.001:
round(.02)
#> [1] 0Created on 2021-10-16 by the reprex package (v2.0.1)
Another, very small point: p-vals are not zero padded, right? With digits.p set to two you get .065, but also .01 instead of the .010 some might prefer.
Do you have a reproducible example? According to the code, the formatting should be 0.0100 when digits.p = 4.
This function simply works with round(), that's why you get 0, and 0 is formatted a <0.001:
Understood. Is that behavior intended though? I bet, there's a reason you did not use signif.
Do you have a reproducible example? According to the code, the formatting should be 0.0100 when digits.p = 4.
I don't because you are right and I was wrong. I did not figure out the rounding thing and got mixed up.
I bet, there's a reason you did not use signif.
Yes, when I started working on sjPlot, I just recently switched from SPSS to R and wasn't aware of that function ;-)
You switched from SPSS and immediately set out to write this gorgeous package? That's a feat.
Are you interested in a MR for that or do you like it just as it is?
Feel free to submit a PR, that would be great!