2

I have 3 different groups of people that completed a language task. Some individuals took a drug and some did not. I would like to investigate the effect of the drug on performance in the different groups. What is the best method to use a Two way ANOVA or ANCOVA?

Ferdi
  • 5,179
A A
  • 21
  • Randomized or not randomized into the three groups? And when did you collect the language score? Just at the end or onc before the drug and one after the drug? Other than the drug, are there any other things that you felt are different between groups? And what are they? – Penguin_Knight Mar 08 '19 at 15:18
  • So the 3 groups were not randomised 1) Depressed 2) Confused 3) controls. The language task was completed after drug consumption . Age was used as a potential confound that may have affected the performance. – A A Mar 09 '19 at 16:06

1 Answers1

1

Techincally you want an ANCOVA if you are including Age as a fixed effect. If you are working in R it can be quite handy to fit this as a linear model (which operates in much the same way), as there are some follow on functions that produce some really helpful graphs. You can do this with:

fit <- lm(LanguageScore ~ Treatment + Age, data = df)

and subsequently check your results using:

 anova(fit)

To take it a step further you can correct for multiple comparisons. A package like predictmeans works really well for this and will even output some descriptive tasks automatically.

predictmeans(fit, modelterm = "Treatment")

The predictmeans function will return a number of graphs but the one which has a least significant differences (LSD) bar is really helpful as it shows you clearly which groups are different from one another without confusing letters, and helps correct for multiple comparisons.

André.B
  • 1,450
  • Thank you so much for this! However, I am using SPSS. Please could you suggest how I should control for multiple comparisons in SPSS? – A A Mar 12 '19 at 08:29
  • I would love to be able to but I am afraid I do not use SPSS. If you are happy to make a foray into R then all you would need to do is run the following code: – André.B Mar 12 '19 at 20:20
  • df <- read.csv(file.choose()) - assuming the file is in .csv format. Then install.packages(c("predictmeans")), followed by library(predictmeans). From here all you need to do is run the code above. Sorry I can't help with SPSS, but you might be able to find some answers if you hunt around online. There are a lot of different corrections - I would recommend either false discovery rate (FDR) or Holm method. – André.B Mar 12 '19 at 20:30