Multiple value Chi Squared test in SPSS

[rwcwilson]

Hello,

I am currently doing a study into factors that affect mortality in a particular type of bacterial infection.

I would like to compare certain patient characteristics. One of which is the specialty that the patient is under. I currently have a variable in SPSS that has three values according to the three specialties I have grouped my patients in: 'Haematological', 'surgical' or 'other'. I would like to see if there is any statistical significance between each of these specialties and whether the patient survived or died during their hospital stay.

I assume because I am dealing with categorical data I have to do a Chi Square test, however I would like to look at whether each specialty is significant individually not just if there is any significant relationship between 'specialties' and death. Is there any way of getting SPSS to do this, or do I need to break them up individually?

 

[Disvengeance]

You can construct odds ratios and 95% confidence intervals for comparisons among the three groups.

 

[rwcwilson]

Hmm, that doesn't seem to work because it says it isn't a 2x2 table

 

[Karabiner]

I would like to look at whether each specialty is significant individually

What do you actually mean by this? You want to know
for each speciality whether the mortality rate is above
a certain level?

With kind regards

K.

 

[rwcwilson]

Basically I would like to know whether there is a statistical difference between the number of people in each of the three specialties in the group that died and those who survived.

 

[Karabiner]

This reads as if you wanted to know whether the distribution
of deaths versus survivals deviates from 50% vs. 50%
[in the population from which your data are sampled]. You
could perform three binomial tests with test value 0.50,
or three one-sample Chi² with expected rates 0.5 versus 0.5.

With kind regards

K.

 

[rwcwilson]

That could be it.

The table I would like to produce is attached.

The continuous variables are relatively simple because I am just going to do a T-test to compare the difference in the means. But for the catagorical variables this becomes slightly complicated. I assume for the catagorical variables where there are only two possibilities e.g. sex, admission to HDU/ITU, required dialysis etc. I can just do a chi-square test.

Does what you suggest work for the Specialty variable? Or is there a better way of doing this.

 

[Karabiner]

You could do a Chi² on the 3x2 specialty x survival table, and,
if this turns out significant, then perform 3 pairwise comparisons
(with adjusted alpha level, e.g. Bonferroni-adjustment).

With kind regards

K.