diff --git a/vignettes/AQuickStart.Rmd b/vignettes/AQuickStart.Rmd index 5b2c77f..b170b4a 100644 --- a/vignettes/AQuickStart.Rmd +++ b/vignettes/AQuickStart.Rmd @@ -50,7 +50,7 @@ when there is more than one factor. You are better off keeping steps 1 and 2 sep What you do in step 2 depends on how many factors you have, and how they relate. ### One-factor model {#one-factor} -If one-factor model fits well and the factor is named `treatment`, do +If a one-factor model fits well and the factor is named `treatment`, do ```r EMM <- emmeans(model, "treatment") # or emmeans(model, ~ treatment) EMM # display the means @@ -139,7 +139,7 @@ return `summary_emm` objects (or lists thereof, class `summary_eml`): ``` SEMM <- summary(EMM) ``` -If you display `EMM` and `SEMM`, they *look* identical; that's because `emmGrid` objects are displayed using `summary()`. But they are not identical. `EMM` has all the ingredients needed to do further analysis, e.g. `contrast(EMM, "consec")` will estimate comparisons between consecutive `Treatment` means. But `SEMM` is just an annotated data frame and we can do no further analysis with it. Similarly, we can change how `EMM` is displayed via arguments to `summary()` or relatives, whil;e in `SEMM`, everything has been computed and those results are locked-in. +If you display `EMM` and `SEMM`, they *look* identical; that's because `emmGrid` objects are displayed using `summary()`. But they are not identical. `EMM` has all the ingredients needed to do further analysis, e.g. `contrast(EMM, "consec")` will estimate comparisons between consecutive `Treatment` means. But `SEMM` is just an annotated data frame and we can do no further analysis with it. Similarly, we can change how `EMM` is displayed via arguments to `summary()` or relatives, while in `SEMM`, everything has been computed and those results are locked-in. ## Common things that can go wrong {#problems} @@ -178,7 +178,7 @@ The `pairwise ~` construct is generally useful if you have only one factor; otherwise, it likely gives you results you don't want. ## Further reading {#more} -There are several of these vignettes that offser more details and +There are several of these vignettes that offer more details and more advanced topics. [An index of all these vignette topics is available here](vignette-topics.html). The strings linked below are the names of the vignettes; i.e., they can diff --git a/vignettes/basics.Rmd b/vignettes/basics.Rmd index b18490e..d67314f 100644 --- a/vignettes/basics.Rmd +++ b/vignettes/basics.Rmd @@ -240,7 +240,7 @@ ref_grid(mod5) The reference grid for `mod5` is different from that for `mod4` because in those models, `percent` is a factor in `mod4` and a covariate in `mod5`. It is possible to modify the reference grid. In the context of the present example, -it might be inetersting to compare EMMs based on `mod4` and `mod5`, and we can put +it might be interesting to compare EMMs based on `mod4` and `mod5`, and we can put them on an equal footing by using the same `percent` values as reference levels: ```{r} (RG5 <- ref_grid(mod5, at = list(percent = c(9, 12, 15, 18)))) diff --git a/vignettes/comparisons.Rmd b/vignettes/comparisons.Rmd index 2613e77..6d69a24 100644 --- a/vignettes/comparisons.Rmd +++ b/vignettes/comparisons.Rmd @@ -65,14 +65,14 @@ and the differences in the lower triangle. Options exist to switch off any one o and to switch which triangle is used for the latter two. Also, optional arguments are passed. For instance, we can reverse the direction of the comparisons, suppress the display of EMMs, swap where the $P$ values go, -and perform noninferiority tests with a threshold of 0.05 as follows: +and perform non-inferiority tests with a threshold of 0.05 as follows: ```{r} pwpm(pigs.emm.s, means = FALSE, flip = TRUE, # args for pwpm() reverse = TRUE, # args for pairs() side = ">", delta = 0.05, adjust = "none") # args for test() ``` With all three *P* values so small, we have fish, soy, and skim in increasing order of -noninferiority based on the given threshold. +non-inferiority based on the given threshold. When more than one factor is present, an existing or newly specified `by` variables() can split the results into l list of matrices. diff --git a/vignettes/confidence-intervals.Rmd b/vignettes/confidence-intervals.Rmd index 5e17663..a963c45 100644 --- a/vignettes/confidence-intervals.Rmd +++ b/vignettes/confidence-intervals.Rmd @@ -281,10 +281,10 @@ something akin to a Type II analysis of variance. See the [messy-data vignette]( ## Testing equivalence, noninferiority, and nonsuperiority {#equiv} - + The `delta` argument in `summary()` or `test()` allows the user to -specify a threshold value to use in a test of equivalence, noninferiority, -or nonsuperiority. An equivalence test is kind of a backwards significance +specify a threshold value to use in a test of equivalence, non-inferiority, +or non-superiority. An equivalence test is kind of a backwards significance test, where small *P* values are associated with small differences relative to a specified threshold value `delta`. The help page for `summary.emmGrid` gives the details of diff --git a/vignettes/models.Rmd b/vignettes/models.Rmd index 1d5a5c4..9f7a5c3 100644 --- a/vignettes/models.Rmd +++ b/vignettes/models.Rmd @@ -382,7 +382,7 @@ should match `"prob"` or `"latent"`. With `mode = "prob"`, the reference-grid predictions consist of the estimated multinomial probabilities -- and this implies a re-gridding so no link functions are passed on. The `"latent"` mode returns the linear predictor, -recentered so that it averages to zero over the levels of the response +re-centered so that it averages to zero over the levels of the response variable (similar to sum-to-zero contrasts). Thus each latent variable can be regarded as the log probability at that level minus the average log probability over all levels. diff --git a/vignettes/transformations.Rmd b/vignettes/transformations.Rmd index cc9d30d..0ea56b3 100644 --- a/vignettes/transformations.Rmd +++ b/vignettes/transformations.Rmd @@ -434,7 +434,7 @@ The test statistics and P values differ somewhat from those for the odds ratios ###### {#not-logit} We were able to obtain both odds ratios and risk ratios for `neuralgia.glm`. But what if we had not used the logit link? Then the odds ratios would not @@ -643,7 +643,7 @@ emmeans(ismod, "spray", type = "response", bias.adj = TRUE) you will get exactly the same results, plus a warning message that says bias adjustment was disabled. Why? Because in an ordinary GLM like this, we are *already* modeling the mean counts, and the link function is not a response transformation as such, just a part of the relationship -we are specifying between the linear predictor and the mean. Given the simple structure of this dataset, we can verify this by noting that the estimates we have correspond examply to the simple observed mean counts: +we are specifying between the linear predictor and the mean. Given the simple structure of this dataset, we can verify this by noting that the estimates we have correspond exactly to the simple observed mean counts: ```{r} with(InsectSprays, tapply(count, spray, mean)) ``` diff --git a/vignettes/vignette-topics.Rmd b/vignettes/vignette-topics.Rmd index 31c0dff..a848afd 100644 --- a/vignettes/vignette-topics.Rmd +++ b/vignettes/vignette-topics.Rmd @@ -606,7 +606,7 @@ vignette: > * [Registering `recover_data` and `emm_basis` methods](xtending.html#exporting) * [`regrid` argument](transformations.html#stdize) * [`regrid()`](transformations.html#regrid) - * [probit or other lon-logit models](transformations.html#not-logit) + * [probit or other non-logit models](transformations.html#not-logit) * [`regrid` vs. `type`](transformations.html#regrid2) * [to obtain risk ratios](transformations.html#riskrats) * [`transform = "log"`](transformations.html#logs) @@ -690,7 +690,7 @@ vignette: > * [`joint = TRUE`](confidence-intervals.html#joint) * Tests * [Equivalence](confidence-intervals.html#equiv) - * [Noninferiority](confidence-intervals.html#equiv) + * [Non-inferiority](confidence-intervals.html#equiv) * [Nonzero null](confidence-intervals.html#summary) * [One- and two-sided](confidence-intervals.html#summary) * [Too few means](AQuickStart.html#covar) diff --git a/vignettes/xtending.Rmd b/vignettes/xtending.Rmd index a4eef0f..2de3c4d 100644 --- a/vignettes/xtending.Rmd +++ b/vignettes/xtending.Rmd @@ -244,7 +244,7 @@ and `V` is the covariance matrix of those predictions. In those cases, we recomm setting `misc$regrid.flag = TRUE`. Currently, this flag is used only for checking whether the `nuisance` argument can be used in `ref_grid()`, and it is not absolutely necessary because we also check to see if `X` is the identity. But -it provides a more efficient and reliable check. The code for nuisamce factors relies +it provides a more efficient and reliable check. The code for nuisance factors relies on the structure of model matrices where columns are associated with model terms. So it is not possible to process nuisance factors with a re-gridded basis.