topic Re: Choose the correct Statistica test in Statistical Procedures

Choose the correct Statistica test

Ana2017 — Tue, 14 Nov 2017 13:21:39 GMT

To choose the type of statistical analysis I need, these are some point to take into account:

I need a test to address the question is there a difference between groups? (independent groups.) Comparing between groups.

10 collection time points.

Independent variables: Categorical data (nominal).

Dependent variables (treatment): Categorical data (nominal).

4 Treatments.

Data with normal distribution (parametric tests)

Can be the Chi-Square X2 test suitable to be used in this situation?

All the best.

Re: Choose the correct Statistica test

ballardw — Tue, 14 Nov 2017 15:08:42 GMT

What between the groups are you attempting to determine is different between the groups? The mean value? The distribution of values?

What defines a group?

How many categorical variables? How many levels of each categorical variable? and how big is your sample?

Use of the terminology dependent / independent looks more like some sort of regression.

And please describe how you get "data with normal distribution" when apparently everything you have is categorical.

Re: Choose the correct Statistica test

Ana2017 — Tue, 14 Nov 2017 15:24:02 GMT

What between the groups are you attempting to determine is different between the groups? The mean value? The distribution of values?

I want to know it the type of environment affect the animal behaviour. For example, if Enriched environment 1 makes animals eat more often.

What defines a group?

A group is represented by a group of 10 animals in a same environment (Enriched environment 1, Enriched environment 2, Enriched environment 3, or Enriched environment 4). That is why I have 4 groups.

How many categorical variables? How many levels of each categorical variable? and how big is your sample?

Dependent variables: Categorical data (nominal): 3 different animal behaviours (Sleep, eat, drink).

Independent variables (treatment): Categorical data (nominal). 4 different environments (Enriched environment 1, Enriched environment 2, Enriched environment 3, Enriched environment 4).

Sample of 20 animals, 5 in each treatment.

Every 10 minutes a data was taken from each animal (Sleep, eat or drink). So, I have the data of every animal every 10 min.

And please describe how you get "data with normal distribution" when apparently everything you have is categorical. Uppps! Sorry for this misinterpretation. I am wrong in this.

Thank you so much.

Re: Choose the correct Statistica test

art297 — Tue, 14 Nov 2017 15:36:34 GMT

Dependent variables: Categorical data (nominal): 3 different animal behaviours (Sleep, eat, drink).

But you ask do animals sleep/eat/drink more.

What are the four levels of your dependent measures? Possibly, rather than nominal, they may be at least interval appearing.

Art, CEO, AnalystFinder.com

Re: Choose the correct Statistica test

Ana2017 — Tue, 14 Nov 2017 15:46:27 GMT

Dependent variables: Categorical data (nominal): 3 different animal behaviours (Sleep, eat, drink).

But you ask do animals sleep/eat/drink more.

What I am wondering is:

Does the environment has an effect over animal behaviour?

As an example: Does the enviroment 1 makes animales sleep more?

What are the four levels of your dependent measures? Possibly, rather than nominal, they may be at least interval appearing.

mmm... I think that they are nominal instead of interval. The level are different environments, similar if I have used differents types of diet. What makes your take these as intervals?

Thanks.

Re: Choose the correct Statistica test

art297 — Tue, 14 Nov 2017 15:55:50 GMT

You didn't answer the question:

For sleep, what are the four possible values and what does each value represent?

For eat, what are the four possible values and what does each value represent?

For drink, what are the four possible values and what does each value represent?

Art, CEO, AnalystFinder.com

Re: Choose the correct Statistica test

Ana2017 — Tue, 14 Nov 2017 16:29:51 GMT

Sorry, I must said that:

dependet variable is animal behaviour: Classified in 4 cathegories: Sleep, Eat and Drink.

During 3 hours animals were analysed, the activity of each animal was register every 10 min (during that exactly moment) and classified into Sleep, Eat or Drink.

Moment Animal Treatment Behaviour

0 1 1 Eat

10' 1 1 Sleep

20' 1 1 Sleep

. . . .

(0 to 180´) (1 to 20) (1 to 4) (Eat, Sleep or Drink)

Thanks.

Re: Choose the correct Statistica test

Ana2017 — Tue, 14 Nov 2017 16:39:56 GMT

Sorry, I did not explain it clear.

The dependent variable is: Animal behaviour, which consist in 3 different activities: Eat, Sleep, and Drink.

During a period of 3 hours, I registered the activity (Eat, sleep or drink) of each animal every 10 min.

Time Animal Treatment Behaviour

0 1 1 Eat

10’ 1 1 Sleep

20’ 1 1 Sleep

. . . .

(from 0 to 180´) (1 to 20) (1 to 4) (Eat, Sleep or Drink)

Thanks.

Re: Choose the correct Statistica test

art297 — Tue, 14 Nov 2017 17:23:56 GMT

One possibility would be to write some code to restructure your data so that there will only be one record per animal, with each record containing animal number, treatment, and three measures, namely percent of time eating, percent of time sleeping and percent of time drinking.

If you do that, then you will have three interval measures and could use ANOVA, MANOVA or other parametric test.

Art, CEO, AnalystFinder.com

Re: Choose the correct Statistica test

Ana2017 — Tue, 14 Nov 2017 17:37:43 GMT

Do you mean like this...

Animal Treatment % Eat % Drink %Sleep

1 1 30 20 50

2 1 10 20 70

3 1 60 20 20

4 1 05 10 85

5 1 30 20 50

6 2 70 20 10

7 2 30 20 50

(from 1 to 20) (1 to 4) (Eat + Sleep + Drink= 100)

Can you share some Procedure (ANOVA or MANOVA or something) suitable for me to use, please?

Thanks.

Re: Choose the correct Statistica test

Ana2017 — Tue, 14 Nov 2017 17:45:53 GMT

Do you mean this…?

Animal Treatment % Eat % Drink % Sleep

1 1 20 30 50

2 1 20 30 50

3 1 40 10 50

4 1 20 70 10

5 1 10 10 80

6 2 20 30 50

7 2 60 20 20

8 2 20 30 50

(from 1 to 20) (1 to 20) (%Eat + %Sleep + %Drink = 100)

Thanks.

CAn you suggest me any ANOVA or MANOVA procedure to use? Thanks.

Re: Choose the correct Statistica test

art297 — Tue, 14 Nov 2017 18:03:37 GMT

Yes, that is what I was thinking, but you have to provide some more specific info. How many animals are there, how many treatments are there, and does each animal get only one treatment, multiple treatments, or all treatments.

Art, CEO, AnalystFinder.com

Re: Choose the correct Statistica test

Ana2017 — Tue, 14 Nov 2017 18:06:56 GMT

There are 20 animals (5 per treatment).
There 4 Treatments.
Each animal get only one treatment.

Re: Choose the correct Statistica test

art297 — Tue, 14 Nov 2017 18:46:39 GMT

One of the forum's statisticians will have to chime in but, from my non-statistician perspective, I'd guess that your animals per treatment may be too small to have sufficient power to be able to reject a null hypothesis.

You just barely meet the n-k (i.e., 5-4 > 0) condition, but you'd have to run the tests (and probably power estimates) to be sure.

As for the code regarding an ANOVA,

proc anova data=yourdatasetname;
  class treatment;
  model sleep = treatment;
run;

proc anova data=yourdatasetname;
  class treatment;
  model eat = treatment;
run;

proc anova data=yourdatasetname;
  class treatment;
  model drink = treatment;
run;

Similarly, for the MANOVA,

proc glm data = yourdatasetname;
  class treatment;
  model eat drink sleep = treatment / SS3;
  manova h = treatment;
run;

Art, CEO, AnalystFinder.com

Re: Choose the correct Statistica test

sld — Tue, 14 Nov 2017 21:40:40 GMT

Assuming that you have no hypotheses to test about behaviour over the 3 hour period (in other words, time is not a factor in the model), I would consider a generalized linear model with a multinomial distribution and a generalized logit link, where Behaviour (with 3 outcomes) is the multinomial response.

Unless you are familiar with this type of model, it probably is not a good place to start due to its complexity. So you could scale back, and look at a generalized linear model with a binomial distribution for each Behaviour separately. For example, you could compare the proportion of Eat observations among treatments using (number of Eat observations during 3 hour period)/(total number of observations during 3 hour period) as the binomial response. If all went well, you could then ramp up to a multinomial response.

Re: Choose the correct Statistica test

Ana2017 — Fri, 17 Nov 2017 14:44:34 GMT

Dear:

I followed your suggestion and I found significance differences between treatments in each activity. Using both ANOVA and MANOVA.

P-values:

Eat: 0 . 0 0 0 1

Drink: 0. 0 3 9 7

Sleep: <0 . 0 0 0 1

Now, how can I find the differences between treatments? Because for example, for the activity EAT maybe there is a difference between treatment 4 and 2, but not between treatment 4 and 3.

Re: Choose the correct Statistica test

Ana2017 — Fri, 17 Nov 2017 14:58:17 GMT

Dear:

I followed your suggestion and I found significance differences between treatments in each activity. Using both ANOVA and MANOVA.

Eat: 0 . 0 0 0 1

Drink: 0. 0 3 9 7

Sleep: <0 . 0 0 0 1

Now, how can I find the differences between treatments? Because for example, for eat maybe there is a difference between treatment 4 and 2, but not between treatment 4 and 3.

Re: Choose the correct Statistica test

StatDave — Fri, 17 Nov 2017 15:07:14 GMT

As sld mentioned, the response is multinomial - it is an unordered, categorical response. Further, you have repeated measures per animal over time. An appropriate model can be fit using PROC GEE using the DIST=MULT and LINK=GLOGIT options in the MODEL statement and including a REPEATED statement. It might look like this if you want Treatment as the only predictor. Assuming you want comparisons among the treatments, the LSMEANS statement provides that.

proc gee;
class treatment animal;
model behavior = treatment / dist=mult link=glogit;
repeated subject=animal;
lsmeans treatment / diff;
run;

Re: Choose the correct Statistica test

Ana2017 — Fri, 17 Nov 2017 15:52:36 GMT

Hi:

Do you meant this way? Changing in the place of behaviour, each activity one by one (eat, drink and sleep)?

proc gee;
class treatment animal;
model eat= treatment / dist=mult link=glogit;
repeated subject=animal;
lsmeans treatment / diff;
run;

proc gee;
class treatment animal;
model drink = treatment / dist=mult link=glogit;
repeated subject=animal;
lsmeans treatment / diff;
run;

proc gee;
class treatment animal;
model sleep = treatment / dist=mult link=glogit;
repeated subject=animal;
lsmeans treatment / diff;
run;

I got this error.... please look...

ERROR: Only the Independent working correlation type is available for multinomial models.

NOTE: The SAS System stopped processing this step because of errors.

NOTE: PROCEDIMIENTO GEE used (Total process time):

Regards.

Re: Choose the correct Statistica test

StatDave — Fri, 17 Nov 2017 15:59:20 GMT

No, I mean you should use the Behaviour variable which has values Eat or Sleep or Drink and to run PROC GEE just once with Behaviour as the response variable as I showed. The default should be the independence correlation structure, so I don't know why you got that error unless you specified the TYPE= (or CORR=) option in REPEATED with a value other than IND. To be sure, specify TYPE=IND in REPEATED:

repeated subject=animal / type=ind;