maybe a logistic regression would be of use here?
Hello,
I'm a grad student looking for help with what is probably a very simple question to someone with the right knowledge. I'm trying to finish my Master's research project which is a cross-sectional study examining prevalence of a disease within a local population. I've done all the descriptive statistics I need, but am I looking for the best inferential test(s) to accept or reject my null hypothesis and sub-hypotheses. I have 211 subjects who either test positive or negative for the disease, and I hypothesized that different sub-populations would be either negative or positive as well. Thus, some of the sample sizes are even smaller depending on the population being examined. I was told chi-square wouldn't work because of my small sample size? Could anyone tell me a better way to test this? (Hypotheses are that my data will reflect other current research results.)
Thanks very much!
maybe a logistic regression would be of use here?
"If you torture the data long enough it will eventually confess."
-Ronald Harry Coase -
What kind of subgroups are you interested in? I.e. what are your independent variables? Will you be using any control variables?
Be careful with terminology. Conventional significance testing doesn't allow us to accept null hypotheses, only to reject them. A non-significant p value only means we haven't got enough evidence to reject the null. It doesn't mean that we have evidence to support or to accept it.
Yes, a non-significant p value only allows for rejection of the Ho.
As far as logistical regression goes, this looks like it would work for me, because I have nominal data variables to be paired with other nominal/ordinal variables. Is it possible to run logistic regression or multiple logistic regression in excel? Is it very difficult?
Thanks!
I don't have emotions and sometimes that makes me very sad.
I'm not sure; you'd probably need some kind of add-in to do so. But it's probably best not to use Excel for serious data analysis (some reasons here).
It would be easier and better to use a proper stats package or computing environment. E.g. R allows you to run a logistic regression very simply using code like this:
Code:glm(Y ~ X1 + X2 + X3, data = yourdata, family = "binomial")
hi,
coming back to the original question, I wonder how the sample size is too low for chi-squared. If I remember correctly the rule of the thumb is 5 data points in each cell which does not seem difficult with your number of samples.
Concerning the logistic regression, if you do not have continuous independent variables there should be no difference to a chi-squared I believe. If you do have continuous X-s then chis-squared is not an option at all.
regards
rogojel
PrinceOfDarkWater (08-05-2013)
Tweet |