added journal references for data sources, improved graph cosmetics

abd17-1.Rmd

@@ -11,6 +11,8 @@ minor: ordinary least squares
 ---
 
 # Data
+The data are from Carré, J. M., and C. M McCormick. 2008. _Proceedings of the
+Royal Society of London, Series B, Biological Sciences_ **275**:2651–2656.
 ```{r input}
 require(rms)
 knitrSet(lang='markdown')
@@ -19,12 +21,13 @@ face    <- c(159,167,165,172,179,177,174,174,177,178,176,
 penalty <- c(44,143,157,14,27,35,85,113,147,151,199,
              106,120,123,80,253,123,110,161,195,295)/100
 d <- data.frame(face, penalty)
-d <- upData(d, labels=c(face    = 'Face Width:Height Ratio',
-                        penalty = 'Penalty Minutes'))
+xl <- 'Face Width:Height Ratio'
+yl <- 'Penalty Minutes'
+d <- upData(d, labels=c(face = xl, penalty = yl))
 ```
 ```{r show}
 d    # same as print(d)
-ggplot(d, aes(x=face, y=penalty)) + geom_point()
+ggplot(d, aes(x=face, y=penalty)) + geom_point() + xlab(xl) + ylab(yl)
 dd <- datadist(d); options(datadist='dd')
 ```
 # Linear Regression Fit
@@ -42,7 +45,7 @@ seb <- sqrt(mse / sxx)
 c(xbar=mean(x), ybar=mean(y), sxx=sxx, syy=syy, sxy=sxy, b=b, mse=mse, seb=seb)
 tcrit <- qt(0.975, n - 2)
 tcrit
-f <- ols(penalty ~ face)
+f <- ols(penalty ~ face, data=d)
 f
 anova(f)
 ```
@@ -53,12 +56,12 @@ r <- resid(f)
 par(mfrow=c(1,2))   # 1x2 matrix of plots
 plot(fitted(f), r); abline(h=0)  # yhat vs. r
 qqnorm(r)           # linearity indicates normality
-qqline(r)
+qqline(as.numeric(r))
 ```
 
 # Partial Effects of Predictors With Raw Data
 ```{r peffects}
-ggplot(Predict(f, face=seq(1.59, 2.07, length=100))) +
+ggplot(Predict(f, face=seq(1.59, 2.07, length=100)), ylab=yl) +
     geom_point(aes(x=face, y=penalty), d)
 ```
 

abd17-20.Rmd

@@ -11,6 +11,7 @@ minor: ordinary least squares
 ---
 
 # Data
+The data are from Heusner, A. A. 1991. Size and power in mammals. _Journal of Experimental Biology_ **160**: 25-54.
 ```{r setup}
 require(rms)
 knitrSet(lang='markdown')
@@ -42,6 +43,8 @@ close(data)
 d <- upData(d, labels=c(species = 'Species', mass='Mass',
                         rate    = 'Basal metabolic rate'),
 							 units=c(mass='g', rate='watts'))
+xl <- label(d$mass, plot=TRUE, units=TRUE)
+yl <- label(d$rate, plot=TRUE, units=TRUE)
 ```
 Raw data are plotted using log10 scales for x and y.
 ```{r show,h=4}
@@ -55,8 +58,8 @@ w <- function(x) {
 	 }
 
 ggplot(d, aes(x=mass, y=rate, color=species)) + geom_point() +
-  geom_text(aes(label=w(species)), size=3, hjust=1.2) + 
-  scale_x_log10() + scale_y_log10() +
+  geom_text(aes(label=w(species)), size=3, hjust=1.3) + 
+  scale_x_log10() + scale_y_log10() + xlab(xl) + ylab(yl) +
   guides(col=guide_legend(nrow=4, override.aes=list(size=1))) +
   theme(legend.position='bottom', legend.text=element_text(size=4), legend.key.size=unit(2, 'mm'))
 dd <- datadist(d); options(datadist='dd')
@@ -68,7 +71,7 @@ With R you can log while fitting, without creating new variables.  A predicted v
 ```{r fit}
 f <- ols(log10(rate) ~ log10(mass), data=d)
 f
-ggplot(Predict(f, mass=seq(50, 70000, by=50), fun=function(y) 10^y)) +
+ggplot(Predict(f, mass=seq(50, 70000, by=50), fun=function(y) 10^y), ylab=yl) +
   geom_point(aes(x=mass, y=rate), data=d)
 ```
 The confidence interval for $\beta$ (slope of mass on the log10 scale) can be obtained from the confidence interval for a 10-fold change in mass.

abd17-28.Rmd

@@ -11,6 +11,7 @@ minor: ordinary least squares
 ---
 
 # Data
+The data are from Sherwood, C. C., _et al_. 2006. _Proceedings of the National Academy of Sciences (USA)_ **103**: 13606–13611.
 ```{r setup}
 require(rms)
 knitrSet(lang='markdown')
@@ -43,6 +44,8 @@ close(data)
 d <- upData(d, labels=c(species = 'Species', mass='Brain Mass',
                         ratio   = 'Glia-neuron Ratio'),
 							 units=c(mass='g'))
+xl <- label(d$mass, plot=TRUE, units=TRUE)
+yl <- label(d$ratio)
 ```
 Raw data are logging x then logging both x and y.
 ```{r show,h=4}
@@ -57,15 +60,15 @@ w <- function(x) {
 	 }
 
 ggplot(d, aes(x=mass, y=ratio, color=species)) + geom_point() +
-  geom_text(aes(label=w(species)), size=2, hjust=1.2) + 
-  scale_x_log10(breaks=c(10,25,50,100,200,500,1000)) +
+  geom_text(aes(label=w(species)), size=2, hjust=1.3) + 
+  scale_x_log10(breaks=c(10,25,50,100,200,500,1000)) + xlab(xl) + ylab(yl) +
   guides(col=guide_legend(nrow=4, override.aes=list(size=1))) +
   theme(legend.position='bottom', legend.text=element_text(size=4),
 	      legend.key.size=unit(2, 'mm'))
 
 ggplot(d, aes(x=mass, y=ratio, color=species)) + geom_point() +
-  geom_text(aes(label=w(species)), size=2, hjust=1.2) + 
-  scale_x_log10(breaks=c(10,25,50,100,200,500,1000)) +
+  geom_text(aes(label=w(species)), size=2, hjust=1.3) + 
+  scale_x_log10(breaks=c(10,25,50,100,200,500,1000)) + xlab(xl) + ylab(yl) +
 	scale_y_log10(breaks=c(.25,.5,.75,1,1.25,1.5,2)) +
 	theme(legend.position='none')
 
@@ -77,7 +80,7 @@ Note that it is unusual not to log a ratio, but the preliminary analysis in ABD
 ```{r fit}
 f <- ols(ratio ~ log(mass), data=d, subset=species != 'homo sapiens')
 f
-ggplot(Predict(f, mass=seq(10, 1400, by=25))) +
+ggplot(Predict(f, mass=seq(10, 1400, by=25)), ylab=yl) +
   geom_point(aes(x=mass, y=ratio), data=d)
 ```