# Thread: Does ANOVA tell which group is more /less than the other , or just they are different

1. ## Does ANOVA tell which group is more /less than the other , or just they are different

I was reading a peer reviewed paper titled: (Going hybrid: An analysis of consumer purchase motivations) where the authors used factor analysis (principal factor analysis) to know dimensionality of hybrid vehicles purchase motivations. As part of the methodology section of the paper, authors stated:

"For both factor analyses, we conducted an Analysis of Variance (ANOVA) to check the effect of previous and other car ownerships (previous hybrid ownership, previous and current ownerships of cars manufactured by Toyota or other manufacturers) and geographic location (within the M25 – an orbital motor way encircling Greater London – vs. outside the M25). "

So authors wanted to check if there is any effect of car ownership and geographical location , then, authors stated the following in results section:
“Both factors are affected by location. For those living within the M25, the perceived practical benefits are more likely to act as motivations, particularly the policy related benefits of Factor 4 (F=25.139; p=.000). The effect of location for Factor 5 is less strong, but still significant (F=8.897; p=.000).”

my question: ANOVA is used to check if there is any difference between groups means or not, but Does it tell which group is more or less? Is it scientifically correct to use ANOVA to say “For those living within the M25, the perceived practical benefits are more likely to act as motivations”?

Thank you so much for taking the time to answer me

Regards

2. ## Re: Does ANOVA tell which group is more /less than the other , or just they are diffe

I would say that one needs to be more specific when using the term ANOVA. Specifically a t-test is an ANOVA; a two-independent samples t-test is an ANOVA; a Repeated Measures design is an ANOVA; etc. Also, one could say ANOVA is subsumed under Regression Analysis e.g. a One-Way ANOVA that tests for differences between means could be conducted by constructing a Regression model with dummy, effect, or contrast coding schemes.

3. ## Re: Does ANOVA tell which group is more /less than the other , or just they are diffe

Thank you. Exactly, as I am critiquing the paper I mentioned, I will put a note on how authors were not really specific in stating the methods used. Also, there were not explicitly stating their hypothesis.

 Tweet