# Which test to use?

#### bsing82

##### New Member
Hello guys,

I am conducting a study to evaluate the effect of supplementation on serum vitamin levels. There are four categories based on the levels- a, b, c and d and subjects in each have received supplementation based on these levels. Category c and d both receive the same dose. At the end of the study period the levels have to be rechecked and reclassified into a, b, c and d categories.

Can anyone tell me what would be the most appropriate test to determine the p value? One-way ANOVA or something else?

#### Karabiner

##### TS Contributor
There are four categories based on the levels- a, b, c and d
Are these really 4 categories, or is it an ordinal
variable (like "small, medium, high, very high")?

With kind regards

K.

#### PeterFlom

##### New Member
Sounds to me like the levels are measurements of a dose, which could be considered ratio data rather than categorical data. You could thus use regression and consider dose as a continuous variable. You'd first want to plot things to look at possible shapes of the relationship. You'd also want to incorporate any substantive knowledge of the dose-response relationship. As a check, you could also run the same regression with dose as a categorical variable and then compare results.

#### seanw

##### New Member
If category c and d receive the same dose, then category can't be dose. They must be different supplements or something.

#### bsing82

##### New Member
Thank you all for your replies.

Each category represents a level range, e.g. a: 40-50, b: 30-40, etc.

Each category received a dose of the vitamin based on the initial level, e.g. 5 u for category a, 10 u for b, 20 u for c and d.

The levels are to be rechecked in all patients after 2 months of therapy.

#### Karabiner

##### TS Contributor
So you have repeated ordinal measurements.
A sign test could be used here.

With kind regards

K.