What kind of statistic methods?

[question]

Okey, i am confused about this matter.

 

 

I have two different samples. One is 5 untreated/control samples (same kind) and the other is a 50 treated samples (same kind). I wonder what kind of statistic methods i can use to compare the results between them? I don't think that i can use ANOVA, since they need at least 3 different samples.

 

Besides i can not use T-test or Z-test either, since the former needs both samples to be at a small number n<30, while the latter needs both samples to be large n>30. My samples is n=5 (controll samples) and n=50 (treated samples).

 

Hope for inputs.

 

Thank you!

[Glider]

I'm a bit confused too. You say you have 2 samples, but your post also implies you have 55 samples (5 control 50 treated). What do you mean by 'sample'? Give a little more information regarding your experiment and we may be able to help.

[question]

You can consider my problem as this: I have two groups. Group A gets treatment while gropu B is the controll where it does not get treatment. In A i have 50 samples/observations (n=50) and in B i have 5 samples/observations (n=5). So what kind of tests should i use to compare group A versus group B?

 

 

 

Hope for your inputs and ideas!

 

 

Thank you.

[Glider]

Well, what you are testing for determines the test. If you are testing for a difference in means between samples, then you have to use a t-test.

 

The difference in sample sizes won't influence what test you need to use, but as it is such a large difference, it will weaken your experiment. The greatest power is achieved when the sample sizes are equal.

 

What are you measuring? i.e. are the data parametric or non-parametric?

[question]

Well' date=' what you are testing [i']for[/i] determines the test. If you are testing for a difference in means between samples, then you have to use a t-test.

 

The difference in sample sizes won't influence what test you need to use, but as it is such a large difference, it will weaken your experiment. The greatest power is achieved when the sample sizes are equal.

 

What are you measuring? i.e. are the data parametric or non-parametric?

 

Aww, someone adviced me to use U-test mann-Whitney, but i don't know if it is the best. What do you think about it? I am testing for the fold difference of some gene expressions between the treated samples (pituitary tumour samples) relative to the controll samples (pituitary normal samples). As i know the experiment must be non-parametric, since the populations don't have normal distributions. Or is it parametric?

[Glider]

If the data are ordinal, the Mann-Whitney U test is the one to use (non-parametric test of difference between two independent samples).

 

So what is your level of measure? If it is interval or ratio, you can use a t-test. If it is ordinal then use the Mann-Whitney U. If it is nominal, you will have to use a Chi Squared test.