Looping many one-sided ANOVA in R

Question

I would like to run 100 ANOVA for different numeric vectors.

My numeric vectors are (ruy, fjr, akf....) from my data.frame

aa=aov(data.frame$ruy~data.frame$Group)
anova(aa)
ab=aov(data.frame$fjr~data.frame$Group)
anova(ab)
ac=aov(data.frame$akf~data.frame$Group)
anova(ac)
.....
.....

My looping skills are poor so please be nice.

score 2 · Accepted Answer · answered Nov 12 '14 at 14:36

2

Since aov is based on lm you can cbind dependent variables on the LHS, which results in seperate models being run:

formula <- as.formula(paste0("cbind(", paste(names(iris)[-5], collapse = ","), ") ~ Species"))

fit <- aov(formula, data=iris)
summary(fit)
# Response Sepal.Length :
#             Df Sum Sq Mean Sq F value    Pr(>F)    
#Species       2 63.212  31.606  119.26 < 2.2e-16 ***
#Residuals   147 38.956   0.265                      
#---
#Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#
# Response Sepal.Width :
#             Df Sum Sq Mean Sq F value    Pr(>F)    
#Species       2 11.345  5.6725   49.16 < 2.2e-16 ***
#Residuals   147 16.962  0.1154                      
#---
#Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#
#</snip>

answered Nov 12 '14 at 14:36

Roland

127,288
10
191
288

I should have written that I have alot of NA:s and when I do the ANOVA manually R ignores them. I do not want to treat NA as 0. – Einar Nov 12 '14 at 16:26
Both this answer by Roland and the answer below by @Thothal are compelling to my purposes, however I have a question of my own. How would one expand with method to run multiple post hoc tests? The linear model works, but the results of the information are structured such that Post Hoc tests cannot be easily run and "stacked". Any thoughts? – David Weisser May 04 '20 at 20:49
It depends on what kind of post-hoc test you intend to run. Do you just want to adjust the p-values shown in this answer or do you want to run a test with each model? For the latter, use the approach shown by thothal. – Roland May 05 '20 at 04:57

score 1 · Answer 2 · answered Nov 12 '14 at 15:35

Alternatively, you can loop over the responses to create a list where each eleemnt corresponds to one model instead of the excellent answer from Roland which generates a 'single' model with multiple responses. This could be usefule if you want (in a later step) work with the generated models seperately:

responseList <- names(iris)[-5]
modelList    <- lapply(responseList, function(resp) {
                           mF <- formula(paste(resp, " ~ Species"))
                           aov(mF, data = iris)
                })

Then you could use lapply again to run the summary on the aov models:

lapply(modelList, summary)

As mentioned, the solution from Roland gives you a multiple responses model (class(fit)) whereas the solution above gives you a list of (single) response models. Whatever you prefer mainly depends on how you want to work with the result.

Looping many one-sided ANOVA in R

2 Answers2

Linked