fixing models

7_02_two_way_anova_R.Rmd

@@ -237,7 +237,7 @@ We want to plot some sample statistics (means and standard errors) so we first h
 festuca_stats <- 
   festuca %>% 
   group_by(Calluna, pH) %>% # <- remember to group by the two factors
-  summarise(Means = mean(Weight), SEs = sd(Weight)/n())
+  summarise(Means = mean(Weight), SEs = sd(Weight)/sqrt(n()))
 festuca_stats
 ```
 Once we've constructed a data frame containing the descriptive statistics we can make a plot:

README.md

@@ -2,13 +2,16 @@ This is the course book for the Data Analysis in R course in the Department of A
 
 ## Notes for the author
 
+
 *   Do not place files in any directory ending in "_files". This will get moved into "_bookdown_files" and will not be tracked by git.
 
 ## To do list
 
+*   Use th phrase distributional assumptions throughout to refer to normality and homoskedasticity assumptions. Make sure the assumptions are expresssed in terms of the residuals rather than "data".
+
 *   Ask students to build their own data frames containing the data in the t-test exercises.
 
-*   Throw in a large sample t-test that is a little non-normal
+*   Throw in a large sample t-test where the sample disttibution is a little non-normal.
 
 *   Chi-square GOF chapter needs a reminder that if the counts are in a data frame they need to be extracted with `$`
 
@@ -20,8 +23,10 @@ This is the course book for the Data Analysis in R course in the Department of A
 
 *   Make the 'classic checks' part of the regression diagnostics section shorter: remove the code and just show the plot with the data and fitted line. Students seems to think they have to do all this as well as use 'plot' at the moment.
 
-*  Remove rough checking of assumptions from one way anova and make that section about understanding the patterns. It confuses students.
+*  Remove rough checking of assumptions from one way anova and make that section about understanding the patterns. It confuses students -- they're not sure which methoid to use to evaluate assumptions.
 
 *  Remind them that the dignostics are based on the residuals in the anova diagnostics.
 
 *  Change 'dependent' to 'response' throughout, AND use 'dependent' in the anova sections to hammer home the all models are lm's
+
+* Remove all references to 'non-parametric correlation' in the regression/correlation chapter, move Spearman's rho stuff to the non-parametric tests chapter (and rememebr to revise the exercises)

data_csv/LEAF_DAMAGE.CSV

@@ -1 +1,41 @@
-ID,LeafType,Damage1,Damaged,652,Damaged,03,Damaged,104,Damaged,455,Damaged,556,Damaged,457,Damaged,08,Damaged,109,Damaged,510,Damaged,7011,Damaged,5512,Damaged,4013,Damaged,014,Damaged,7015,Damaged,6516,Damaged,3517,Damaged,6518,Damaged,4519,Damaged,020,Damaged,201,Notdamaged,52,Notdamaged,253,Notdamaged,304,Notdamaged,405,Notdamaged,156,Notdamaged,57,Notdamaged,358,Notdamaged,459,Notdamaged,3010,Notdamaged,1011,Notdamaged,1512,Notdamaged,5013,Notdamaged,014,Notdamaged,015,Notdamaged,516,Notdamaged,4517,Notdamaged,2018,Notdamaged,1019,Notdamaged,3020,Notdamaged,0
+ID,LeafType,Damage
+1,Notdamaged,65
+2,Notdamaged,0
+3,Notdamaged,10
+4,Notdamaged,45
+5,Notdamaged,55
+6,Notdamaged,45
+7,Notdamaged,0
+8,Notdamaged,10
+9,Notdamaged,5
+10,Notdamaged,70
+11,Notdamaged,55
+12,Notdamaged,40
+13,Notdamaged,0
+14,Notdamaged,70
+15,Notdamaged,65
+16,Notdamaged,35
+17,Notdamaged,65
+18,Notdamaged,45
+19,Notdamaged,0
+20,Notdamaged,20
+1,Damaged,5
+2,Damaged,25
+3,Damaged,30
+4,Damaged,40
+5,Damaged,15
+6,Damaged,5
+7,Damaged,35
+8,Damaged,45
+9,Damaged,30
+10,Damaged,10
+11,Damaged,15
+12,Damaged,50
+13,Damaged,0
+14,Damaged,0
+15,Damaged,5
+16,Damaged,45
+17,Damaged,20
+18,Damaged,10
+19,Damaged,30
+20,Damaged,0