What statistical analysis to preform - ANOVA?

#1
I thought I needed one stat but now am doubting myself so wanting some advice. I originally thought i would run a multi variance ANOVA however, after running a normality test, my data is not normally distributed. Is this ANOVA still ok to use or is there a better stats test to use?

My data is food vs animal ingestion and the data is :
food data - food size (1 or 2), food colour (red, green, black, brown), food fouled (fouled or not fouled) and number ingested
Organism data - organism (1 2 or 3).
I'm looking at various hypothesis but mostly looking at differences i.e did organism 1 eat more of 1 colour and of a specific size. OR in one graph I have a graph showing all the organisms and whether they eat more of one colour of size size and if it was fouled or not.

I'm open to different types of stat tests to use so please make a suggestion if you think ANOVA isn't the best but please explain why. There is lot son information only but a lot is conflicting so i've gotten so confused and now have hit a brick wall.

Any questions please ask and thanks for any help anyone can give.
 

Karabiner

TS Contributor
#2
I thought I needed one stat but now am doubting myself so wanting some advice. I originally thought i would run a multi variance ANOVA however, after running a normality test, my data is not normally distributed.
ANOVAs do not assume normally distributed dependent variables, and normality tests are
practically useless anyway (either too sensitive or too insensitive, depending on sample size).

I'm open to different types of stat tests to use so please make a suggestion if you think ANOVA isn't the best but please explain why.
So you could perhaps explain your study first. What is the topic, what are the research questions?
How were the data collected - for example, was it an experimental study? Did you have 2x4x2x3=
48 independent groups, or were there repeated-measures factors (such as, to the same subjects
were presented fouled food and afterwards non-foul food)? How large is the sample size?

With kind regards

Karabiner
 

Karabiner

TS Contributor
#4
Oh, now I see. Seemingly, all the food data are dependent variables?
Still not sure whether these are categorical (did a subject eat size 1, or
size 2, or both) or continuous (did a subject eat 15 gram from size 1
and 12 gram from size 2).

With kind regards

Karabiner
 
#5
Sorry for not adding more detail before. My study is looking at fish ingestion of food based on various factors. I'm looking at wether fish species at the food and wether or not it had preference to : size, colour or fouled . To do this I had 5 replicates and exposed these replicates to the four colours and one size and one fouled(or not fouled). E.g fish species 1 had 5 fish exposed to all 4 colours of size 1 which was fouled. I then also exposed 5 species to all four colours of the of size 1 and wasn't fouled. This was repeated for all 3 species. I also then looked at wether being together effected the species ingestion of each food and so I did - all three species in one tank and I exposed them to 4 colours of size 1 which was fouled (same process as above but just all species in one tank). I hope that gives more insight
 

Karabiner

TS Contributor
#6
E.g fish species 1 had 5 fish exposed to all 4 colours of size 1 which was fouled. I then also exposed 5 species to all four colours of the of size 1 and wasn't fouled.
What does "e.g" indicate? Does it indicate that you performed more experimental variations for each species,
not just those 2 described?

On which level was your dependent variable measured, was it total food intake by 5 fish (group level), or was it
food intake per each fish (individual level)?

With kind regards

Karabiner
 
Last edited:
#7
sorry, e.g means for example.

and yes it was per individual fish but as these are replicate fish, the mean will be taken from these 5.

Thank you for all your help with this, it is really appreciated. My supervisor started telling me to run certain tests and to me it didn't make sense and just confused me over it all. They told me to run normality tests and then after that chose the test you need and that an ANOVA might not be the right one but hasn't offered any help in researching or getting to the right one.
 

Karabiner

TS Contributor
#8
Yes, I know what e.g. means, but my question was whether you have more comparisons to be made, beyond those 2 you described.
If I understand you correctely, you have 3 independent groups of fish. Each group was exposed to 2 food sizes and 4 food colours and
2 food qualities. There were 4 sessions (small/foul, small/not foul, large/foul, large/nonfoul). If I am correct, then you could consider a
"mixed" ANOVA with a grouping factor (species, 3 levels) and three repeated-measures factors (colour, size, quality).

With kind regards

Karabiner
 
#9
Ah I see what you mean. The e.g was in reference to repeating this for each fish so yes, like what you replied. You suggest an ANOVA which is what I originally thought but I also found two other tests that I thought might work but wasn't totally sure they apply to my data, do you know about Tukey hsd test or a monte carlo test? Even though it may not be suitable for my data it will be great to learn a bit more about these tests just for future knowledge.

Thanks again for all this help,

George