0
$\begingroup$

Before we begin, my knowledge around anything "statistics" related is minimal so any help is appreciated.

I am studying whether performing a surgery at night is safe vs during the day.

I would like to first check whether comparing continuous data (presented as medians) between two groups should be compared with Mann-Whitney U test, and whether categorical data (presented as count or percentage) should be compared with chi-square.

Next, I want to use separate regression analyses to identify independent predictors of an outcome (e.g. surgical complication, length of hospital stay etc). Which regressions do I run for the following:

1 My dependent variable (or outcome) is categorical (yes/no), and my independent variables are categorical or continuous.

2 My dependent variable is continuous, and my independent variables are categorical or continuous.

I also want to report the Odds ratio (and 95% CI) of each predictor on the chosen outcome, to highlight which are most likely to increase/decrease/or have no effect on the outcome.

P.S. I am using SPSS

Thankyou in advance for your help!!!

$\endgroup$
  • $\begingroup$ Please indicate if this is homework or self-assigned work. $\endgroup$ – cherub May 4 '18 at 13:31
  • $\begingroup$ This is self-assigned unfunded research. $\endgroup$ – T. Hall May 5 '18 at 2:19
0
$\begingroup$

First, I am guessing from your question that you are a surgeon or other doctor doing some research. How would you respond to someone asking something like "My knowledge of surgery is minimal, so any help would be appreciated. How do I operate on the heart?"

So, my first suggestion to you is to hire a statistician/data analyst.

Now, if you ignore that advice, either because of cost or some other reason, I'll say this: You have asked about surgical complications and length of hospital stay. Then you asked about categorical dependent variables (yes/no) and continuous dependent variables. Unfortunately, the first two don't fit well into the latter two: For surgical complication you probably want (at least to start) a multinomial logistic regression. For the second, you probably want some kind of survival analysis (although this will depend on other characteristics of your study). For yes/no DVs you usually want logistic regression. For continuous DVs the usual start is linear regression.

$\endgroup$
  • $\begingroup$ Hi Peter, thanks for your help. Perhaps I should try and clarify things a bit better... $\endgroup$ – T. Hall May 4 '18 at 12:04
  • $\begingroup$ For surgical complications I have reported this as either yes (there was a complication) or no. I then want to identify what other variables that I have collected are associated with this outcome (a surgical complication). These variables include time of surgery (night vs day), age in years etc. So the outcome of interest is yes/no, but the variables may be yes/no (categorical?) or age (continuous?). Would logistic regression be the way to go here? Also, is it possible to calculate odds ratios using logistic regression? Many thanks! $\endgroup$ – T. Hall May 4 '18 at 12:11
  • 1
    $\begingroup$ Yes, you can have both categorical and continuous predictors in logistic regression. $\endgroup$ – dankernler May 4 '18 at 13:11
  • $\begingroup$ Yes, you can do that with logistic regression, but making all surgical complications the same ("yes") is probably not ideal. $\endgroup$ – Peter Flom May 4 '18 at 20:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.