Comparing 2 groups with unequal number of samples + comparing multiple groups

zaf

New Member
#1
Hello everyone, this is my first post here. I have been doing some surfing and reading on this forum and I must say, you guys are great.

I am going to take a statistical course in the next couple of months, but unfortunately my questions cannot wait till then as I am analyzing my data now.

Questions A:
2 groups. Group 1: n = 180. Group 2: n= 25 (I think this can be classified as categorical, e.g. distribution of gender in each group).

For this analysis I have used Chi Square. Do you agree? Since it is a 2x2 categorical data)

Question B:
5 groups.
Group1: n = 178 (e.g. 170 +ve, 8 -ve)
Group2: n= 15 (e.g 5 +ve, 10 -ve)
Group3: n = 15
Group4: n = 8
Group5: n = 24

Each group represent a test result, and we are comparing the test results to see if it correlate with the diagnosis.
e.g. 5 ways of measuring blood glucose levels, we want to see if there are any difference between these groups.
Can I use One way ANOVA?

I very much appreciate the help.

Thanks
 

hlsmith

Omega Contributor
#2
Question 1, yes this seems appropriate given your description. However you need to have at least 5 of each sex in both groups. Thus, if you put the data into a 2x2 table, gender by group, each cell needs at least 5 individuals.
 

hlsmith

Omega Contributor
#3
You need to provide more information about the second question, such as, are these the same patients, and what your abbreviations represent.
 

zaf

New Member
#4
For the second questions

Lets say you have a test called Beta to diagnose Disease called "Q". Beta has 5 result possibilities; A, B, C, D and E.

We have 230 patients tested using Beta. The following results were achieved.
170 had a result of A, 25 had a result of B, 20 had a result of C, 8 had a result of D, and 7 had a result of E.

Out of the 230 patients 160 were known to have the disease Q.

Test result A: 155/160 have disease Q
result B: 5/25
result C: 2/20
result D: 0/8
result E: 1/7

Can I use one way anova to show statistical difference between the groups?
 

zaf

New Member
#5
Question 1, yes this seems appropriate given your description. However you need to have at least 5 of each sex in both groups. Thus, if you put the data into a 2x2 table, gender by group, each cell needs at least 5 individuals.
yes each cell in the 2x2 table has more than 5.

what about T-test for question 1 would it be better?
 

rogojel

TS Contributor
#6
hi,
you have discrete data so neither Anova nor the t test is appropriate as they are for continuous data. Chi-squared could work but you do not have enough values in the last 3 cells. Grouping C D and E into one category might work, but you still only have 3 data points instead of 5 in the grouped cell.

I think Fishers exact test could work best for you
http://udel.edu/~mcdonald/statfishers.html

regards
rogojel