Dear Zhanna,

My answer straightly is to your question (title). If the distribution of your data is normal, use Independent T-test. If not, make use of Mann–Whitney U test.

I am a bit confused about your outcome variable. Is it categorical or continuous. @Zhanna Baimukanova could you please clarify this.

Multinominal mixed model. Multinominal because your response variable is categorical with 7 levels, and mixed because you have repeated measures.

You can use independent t-test.

However, I saw data you showed.

You should first of all decide weither you need to use parametric or nonparametric independent t-test.

T Test?!

Mann–Whitney?!

I apologize. Sometimes I want to bang my head against the wall at ResearchGate.

I entirely concur with Jochen Wilhelm

@your problem presented is unclear. Would you mind presenting the 7 mutually exclusive categories of behavior?

By using SPSS, you can do the Mann–Whitney test or independent t test. Also, you need determined the type of data to choose the parametric t test or non-parametric.

Dear Zhanna,

your question is too general for your very specific problem (animal behaviour/movement). Appropriate modelling approach for it are Hidden Markov Models (see :

Article Fine-scale movement patterns and behavioral states of gray t...

Article Analysis of animal accelerometer data using hidden Markov models

Article moveHMM: An R package for the statistical modelling of anima...

Article Hidden Markov Models and Animal Behaviour

(Straightforward testing is not preferable in animal movement modelling due to lack of assumptions.)

How have you decoded the observations into 7 behavioral acts? (Spoiler: via HMM it is quite easy:).

Your observations are time series and may have some time patterns (e. g. autocorrelations). Have you checked it? If it is the case then it would be preferable to incorporate the conditions A and B as exogenous variable into the model(s) and compare it with the HMM without any exogenous variable via an information criterion e. g. BIC instead of the testing.

Or alternatively use multinomial mixed model as Jochen Wilhelm has advised if you have already specified the seven states without modelling.

I believe the data meeting normality and linearity assumption will give it the strength to carry out independent t-test, I am assuming those measures you took are continuous in nature. However, if the assumptions are not met then carry out Mann-Whitney U.

The t test is the most basic, when the variance is not known.

Independent samples t-test is appropriate. You can read further about its application.

Independent T-test

You can use T - test

Depends on what you want to do.

I suggest an Independent Samples t-test which compares the means of two independent groups. For more info and examples, you can check this website.

https://www.statisticshowto.com/probability-and-statistics/t-test/

Abolfazl Ghoodjani , seems you can keep on banging your head :(

Jochen Wilhelm Thank you for your attention

Yes, the best solution is the same

I bang my head against the wall.

I think there is another solution.

Escape from ResearchGate.

Every place has its problems. And every medium has the similar property: it makes smart people smarter and dumb people dumber. Despite all the nonsense written in RG, there are also very good contributions and discussions, too.

This one is a no brainer. Consult a statistician.