merge

lectures/Contrasts.rmd

@@ -1,5 +1,6 @@
 ---
 title: "Introduction to contrasts, and using emmeans in R"
+bibliography: "../qmee.bib"
 author: "Ian Dworkin"
 date: "`r format(Sys.time(),'%d %b %Y')`"
 output:
@@ -480,7 +481,7 @@ head(size_dat)
 
 ### A quick visual summary
 
-```{r}
+```{r beeswarm, fig.width = 14, out.width="100%"}
 ggplot(size_dat, 
        aes( y = length, x = selection:sex, col = sex, shape = replicate)) +
   geom_quasirandom(alpha = 0.8, size = 1.4) +
@@ -537,30 +538,34 @@ This is where contrasts become SO useful!
 
 #### Estimated marginal means 
 First, just the estimated means, for each sex, by selective regime
-```{r}
+```{r marg-means-plot, fig.width = 14}
 thorax_emm <- emmeans(mod1_thorax, specs = ~ sex | selection)
 
 thorax_emm
 
+## rotate strip labels:
+## https://stackoverflow.com/questions/40484090/rotate-switched-facet-labels-in-ggplot2-facet-grid
+rot_strips <-   theme_bw() +
+    theme(text = element_text(size = 16),
+          strip.text.y.right = element_text(angle = 0))
+
 plot(thorax_emm,
-     xlab = "model estimates, thorax length, log2 microM") +
-  theme_bw() +
-  theme(text = element_text(size = 16))
+     xlab = "model estimates, thorax length, log2 µm") +
+    rot_strips
 ```
 
 
 #### side note: Back-transforming in emmeans
 Like I mentioned emmeans can recognize the log2 transformation, so if you prefer the measures or plots in $\mu m$ response scale.
 
-```{r}
+```{r fig.width = 14}
 thorax_emm_response <- emmeans(mod1_thorax, specs = ~ sex | selection, type = "response")
 
 thorax_emm_response
 
 plot(thorax_emm_response,
-     xlab = "model estimates, thorax length, microM") +
-  theme_bw() +
-  theme(text = element_text(size = 16))
+     xlab = "model estimates, thorax length, µm") +
+    rot_strips
 ```
 
 
@@ -583,16 +588,15 @@ confint(SSD_contrasts_treatment)
 ```{r}
 plot(SSD_contrasts_treatment) + 
   geom_vline(xintercept = 0, lty = 2, alpha = 0.5) + 
-  labs(x = "sexual size dimorphism") +
-  theme_bw() +
-  theme(text = element_text(size = 16))
+    labs(x = "sexual size dimorphism") +
+    rot_strips
 ```
 
 
 ### The interaction contrast
 But what we really want is to see how dimorphism changes in the selected treatments, relative to the controls. This is the interaction contrast (contrast of contrasts)
 
-```{r}
+```{r fig.width = 14}
 thorax_ssd <- emmeans(mod1_thorax,  pairwise ~ sex*selection) # warning is letting you know these are not of general use. We only do this as we are forming an interaction contrast.
 
 thorax_ssd_contrasts <- contrast(thorax_ssd[[1]], 
@@ -606,9 +610,9 @@ confint(thorax_ssd_contrasts)
 
 plot(thorax_ssd_contrasts) + 
   geom_vline(xintercept = 0, lty = 2, alpha = 0.5) + 
-  labs(x = "change in SSD relative to control lineages", y = "comparison") +
-  theme_bw() +
-  theme(text = element_text(size = 16))
+    labs(x = "change in SSD relative to control lineages", y = "comparison") +
+    theme_bw() + theme(text = element_text(size = 16))
+
 ```
 
 
@@ -628,4 +632,8 @@ Chapter 5 of the book by Irizarry and Love, Data Analysis for the Life Sciences
 overview of contrast codings 
 https://rpubs.com/timflutre/tuto_contrasts
 
+@schadHow2018
+
+https://bbolker.github.io/mixedmodels-misc/notes/contrasts.pdf
 
+## references

qmee.bib

@@ -721,3 +721,19 @@ @article{vanhoveCollinearity2021
 	keywords = {multiple regression, interpreting regression models, regression assumptions},
 	file = {Full Text PDF:/home/bolker/Documents/zotero_new/storage/TKJW5RM6/Vanhove - 2021 - Collinearity isn't a disease that needs curing.pdf:application/pdf},
 }
+
+@article{schadHow2018,
+  title = {How to Capitalize on a Priori Contrasts in Linear (Mixed) Models: {{A}} Tutorial},
+  shorttitle = {How to Capitalize on a Priori Contrasts in Linear (Mixed) Models},
+  author = {Schad, Daniel J. and Hohenstein, Sven and Vasishth, Shravan and Kliegl, Reinhold},
+  year = {2018},
+  month = jul,
+  journal = {arXiv:1807.10451 [stat]},
+  eprint = {1807.10451},
+  primaryclass = {stat},
+  urldate = {2018-09-10},
+  abstract = {Factorial experiments in research on memory, language, and in other areas are often analyzed using analysis of variance (ANOVA). However, for experimental factors with more than two levels, the ANOVA omnibus F-test is not informative about the source of a main effect or interaction. This is unfortunate as researchers typically have specific hypotheses about which condition means differ from each other. A priori contrasts (i.e., comparisons planned before the sample means are known) between specific conditions or combinations of conditions are the appropriate way to represent such hypotheses in the statistical model. Many researchers have pointed out that contrasts should be "tested instead of, rather than as a supplement to, the ordinary `omnibus' F test" (Hayes, 1973, p. 601). In this tutorial, we explain the mathematics underlying different kinds of contrasts (i.e., treatment, sum, repeated, Helmert, and polynomial contrasts), discuss their properties, and demonstrate how they are applied in the R System for Statistical Computing (R Core Team, 2018). In this context, we explain the generalized inverse which is needed to compute the weight coefficients for contrasts that test hypotheses that are not covered by the default set of contrasts. A detailed understanding of contrast coding is crucial for successful and correct specification in linear models (including linear mixed models). Contrasts defined a priori yield far more precise confirmatory tests of experimental hypotheses than standard omnibus F-test.},
+  archiveprefix = {arxiv},
+  keywords = {Statistics - Methodology},
+  file = {/home/bolker/Zotero/storage/WU4VJNID/Schad et al. - 2018 - How to capitalize on a priori contrasts in linear .pdf;/home/bolker/Zotero/storage/YPA8WMCU/1807.html}
+}