# Basic stat question

#### Drmonz

##### New Member
Hi,

I would like to get your opinion what test to use in the following two scenarion

1. I have 100 patients in total and I want to groups (50 men and 50 women) that have done different frequency of imaging (x-ray) and would like to see if there is any difference between gender and the number/frequency of x ray.

All patients have done imaging, some only 1 xray other perhaps 5 xray.

So what I have is a table with 100 rows and two columns

Patient. Number of x ray gender
A 1 female
B 3 male
C 5 female
etc

What statistical test would you use to see if there is any difference between groups in this matter?

Dr gunnar

#### Buckeye

##### Active Member
What if you plot a histogram and/or boxplot to visualize the data per group? Then maybe there is an obvious difference.

#### hlsmith

##### Less is more. Stay pure. Stay poor.
What were the xrays for, any potential co founders (e.g., age, disease severity, baseline health, history related to clinical suspicion, residence difference, primary care provider) or independence issues (e.g., similar and partial clustering of patients in ordering providers or facilities).

I would use Poisson regression.

#### noetsi

##### No cake for spunky
Any general linear model, t-test, regression, ANOVA etc probably would work.

#### Drmonz

##### New Member
What if you plot a histogram and/or boxplot to visualize the data per group? Then maybe there is an obvious difference.
Hmm thanks, no i want a table with p valueS

#### Drmonz

##### New Member
What were the xrays for, any potential co founders (e.g., age, disease severity, baseline health, history related to clinical suspicion, residence difference, primary care provider) or independence issues (e.g., similar and partial clustering of patients in ordering providers or facilities).

I would use Poisson regression.
Yes co founders CAn be present, maybe poison first for univariate then log regression to adjust for co founders?

#### hlsmith

##### Less is more. Stay pure. Stay poor.
A table with pvalues sucks - you need to report an actual estimate. pvalues don't convey information on effect direction or magnitude and are typically misleading. You can control for confounder via Poisson regression. However your sample size is pretty small for doing too much.