1

I am attempting to determine how to test the relationship between a discrete variable (the length of a string of text, my IV) and a categorical variable (binary, a yes or no outcome, my DV) but I'm having trouble figuring out the best methodology. Would it be incorrect to use a one-way ANOVA or is that a valid method?

I had performed an ANOVA and gotten good results (p-value < 0.05 indicating a relationship) but upon further research I'm less confident that the one-way ANOVA has given me good or valid results.

proofs_challenged
  • 123
  • 2
    There is no correlation when categorical variables are involved, so that goes out the window. What is your research question? – user2974951 Jul 20 '23 at 14:05
  • 3
    How is "length of string" a discrete variable? If your DV is binary, then the usual method would be logistic regression. – Peter Flom Jul 20 '23 at 14:06
  • What’s wrong with the usual Pearson correlation? – Dave Jul 20 '23 at 14:39
  • 1
    What exactly do you mean by "testing the relationship", trying to find out if length of string has an influence at all, or something more detailed? – Ute Jul 20 '23 at 15:01
  • @user2974951 research question is what is the strength of the relationship between these two variables. – proofs_challenged Jul 21 '23 at 19:21
  • @PeterFlom wouldn't length of a string be discrete bc you can't have 2.5 words or 2.25 words but rather only 2 words or 3 words in a string? – proofs_challenged Jul 21 '23 at 19:21
  • @Dave won't Pearson correlation only work between on non-categorical vairables? – proofs_challenged Jul 21 '23 at 19:22
  • @Ute exactly, just to find out if the length of the string has an influence on the outcome (i.e. if a user writes a longer string of text, are they more or less likely to indicate a satisfied answer vs a not satisfied) – proofs_challenged Jul 21 '23 at 19:23
  • @proofs_challenged Oops, my bad. I was thing string like yarn. – Peter Flom Jul 21 '23 at 19:28
  • 1
    How big is your sample size, and what is the range of string lengths? – Ute Jul 22 '23 at 08:29

1 Answers1

0

You have given us very little information to build on, so please edit your question so we can know more (sample size? Showing us an excerpt of data, or a link? Plots? ...) but in the meantime ...

For a binary dependent variable and a numerical covariate, logistic regression is my starting point. Search this site ... and see also Correlations with unordered categorical variables, Correlations between continuous and categorical (nominal) variables, Correlation between dichotomous and continuous variable,

kjetil b halvorsen
  • 77,844