My data is looking at a group of 9 patients with laryngeal cancer. I need to calculate whether the trends in my data (gender, smoking status etc.) are statistically significant or not, and give corresponding p values.
I think I need to set a confidence interval on my data, and then do some sort of t-test, however I am unsure how to go about this.
I have attached a sample of my data. I would be grateful for some help that is explained as simply as possible - I have very little statistical knowledge.
Many thanks.
Hi. You have indicated that the independent variables are the trends (gender, being diagnosed with HPV, ec). These are categorical variables.
However, you need to also specify what your dependent variable is in order to carry out any parametric or non-parametric test.
Since all of the respondents have been diagnosed with cancer, the dependent variable isn't specified. Is it the progression of the cancer in stages or is it their physical health measured as a continuous variable? If you specify the dependent variable, one could suggest some statistical tests.
Right now, if this is all you have, then it's better to just report the percentages.
I have a YouTube video on how to choose a statistical test that might be helpful in the future.
https://www.youtube.com/watch?v=F4AKZ9-u_ic
With nine subjects, it's going to be difficult to conclude much. ... In theory, you might be able to say something relative to UK population characteristics. Like, in your sample, gender is equality split, as with the UK population. So gender doesn't appear to be a factor. But if 67% percent of your subjects smoke, relative to UK proportion of, say, 13% that might be suggestive. But for this, you'd have to adjust for age of your subjects. (Because maybe 67% of UK people aged whatever smoke (?) Or from I see, in the UK, the location, by local government district, may matter a lot with smoking rates.) ... With nine subjects it's just super difficult to tease these things out epidemiologically, even with a simple approach.
I don't know any design or statistical test can compare all these variables!. You need to partition variables, then decide how to analyze. Regards.
The test to adopt depends on what you want to achieve, the question is "What is your research question and hypothesis?" Also, your sample size (9) is too small to talk about statistical significance, you might want to consider descriptive analysis and practical significance rather than running statistical tests.
Suppose your sample size is large enough, I would suggest you go for a t-test as the CLT assumption will be met for the large sample size and your data is on a metric scale. You can also run an ANOVA test on each category to test for their significance. All of these tests can be done using R or any other statistical package.