Anova between 4 variables with different names

[katemcguckin]

Hi,

I really need help asap.  I am caught up in this unsystematic study that I did not set up. I am being torn apart by MDs who want this asap. 

Please help me. 

I need to do:

1. check normality bt 4 vars with different names

2. do an ANOVA or appropriate nonparametric test bt those 4 vars. 

The data looks like this.  (Gosh, I wish people would consult prior to doing their own thing without knowing what is needed.

final_no_
pre5case_ post5case_ final5case_ 5case_
Obs pid ppf5 percent percent percent percent group

1 2 1200 66.7 66.7 66.7 55.6 1
2 222 2900 44.4 66.7 66.7 44.4 1
3 36 1600 66.7 88.9 77.8 66.7 1
4 37 2000 33.3 22.2 33.3 55.6 1
5 39 2300 44.4 66.7 55.6 44.4 1
6 42 1900 33.3 55.6 66.7 66.7 1

[katemcguckin]

And I have more questions. I have been up all night. 

Kate McGuckin

[PaigeMiller]

---

@katemcguckin wrote:

I need to do:

1. check normality bt 4 vars with different names

---

Not sure what "bt" means

2. do an ANOVA or appropriate nonparametric test bt those 4 vars. 

The data looks like this.

final_no_
pre5case_ post5case_ final5case_ 5case_
Obs pid ppf5 percent percent percent percent group

1 2 1200 66.7 66.7 66.7 55.6 1
2 222 2900 44.4 66.7 66.7 44.4 1
3 36 1600 66.7 88.9 77.8 66.7 1
4 37 2000 33.3 22.2 33.3 55.6 1
5 39 2300 44.4 66.7 55.6 44.4 1
6 42 1900 33.3 55.6 66.7 66.7 1

In order to do an ANOVA, you need a Y-variable, and categorical X variables. But none of your variables are obviously (to me) categorical, and you haven't clearly stated which of these variables is a Y variable.

So please state, either in SAS language, or just in words, what model it is you want to fit and what statistical tests you want to perform.

[PaigeMiller]

2. do an ANOVA or appropriate nonparametric test bt those 4 vars. 

The data looks like this.

final_no_
pre5case_ post5case_ final5case_ 5case_
Obs pid ppf5 percent percent percent percent group

1 2 1200 66.7 66.7 66.7 55.6 1
2 222 2900 44.4 66.7 66.7 44.4 1
3 36 1600 66.7 88.9 77.8 66.7 1
4 37 2000 33.3 22.2 33.3 55.6 1
5 39 2300 44.4 66.7 55.6 44.4 1
6 42 1900 33.3 55.6 66.7 66.7 1

[katemcguckin]

there is NO Y variable!!!!
That's why I am so confused. He just wants to see if there is a difference
in the scores between tests.


Any ideas? I have been spending more time on this on my jama papers
without pay. and their abstract is due on Monday. and on Sunday, I have
plans, but he will make me work on this.

the study is poorly designed.

[ChrisHemedinger]

Hi @katemcguckin, from what @PaigeMiller says, it sounds like your data (as you've shared it) won't support the test you've been asked to design.  Is it possible that the data need to be transposed?  If you have 4 variables, do the values of each variable represent a different group?  If so, you'll need to rearrange your data so that the table is "long", with a new variable that indicates the group (former variable) for each measure.

[PaigeMiller]

---

@katemcguckin wrote:
there is NO Y variable!!!!
He just wants to see if there is a difference in the scores between tests.

---

I still don't really understand what you mean. "Scores" -- please define. "Tests" -- please define. Which of your variables are scores? Which of your variables are tests?

In fact, it would be a good idea if you took a step back and describe the data, and described how it was collected. Explain the data to me as if I had never seen your problem before, start from scratch.

[Reeza]

Do what you can do and leave it at that. 

It looks like you have pre and post tests. I would do a few things:

1. Graph the data using a histogram and overlay the pre/post to see the distributions. This assumes you have more data than shown obviously. 

2. Create some summary statistics of each variable. 

3. Re-stack the data, so that you have a single variable with another variable that shows the four categories. Then use that to produce a PROC ANOVA. 

Var Value

1    25

1   24

2 24

2  30

3 20

3 54

4 56

4 78

If you could explain a bit more about the data perhaps we could help provide some more suggestions. 

Good Luck.