Help - which test should I use?

[statskid2812]

Hi!
I have some data, and I am not sure which statistical test I should use. The data comes from a survey.
In the survey the respondents (dentists) was presented for two different cases in which the cases differed slightly in how they would carry out an hypothetical case - would they perform surgery or not based upon the patients condition (the answers was categorical (5 in total), but I have transformed them into numbers ranging 0-1, because three of the answers meant no surgery(0) and two of the answers meant surgery(1)).

I want know whether or not the dentists perform more surgeries in one of the cases. I am unsure whether I should use,
- Paired t-test
- Wilcoxon signed rank test
- Or maybe something else.

Feel free to ask any questions.
Hope someone can help me out a bit.

 

[Karabiner]

Did all dentists receive both scenarios? This would require a dependent samples tests for binary outomes: McNemar's test.
Or, were there two separate groups of dentists? That would be 2x2 frequency table for group and outcome, with Chi² test.

By the way, you might perhaps find this useful in the future.

With kind regards

Karabiner

 

[statskid2812]

Did all dentists receive both scenarios? This would require a dependent samples tests for binary outomes: McNemar's test.
Or, were there two separate groups of dentists? That would be 2x2 frequency table for group and outcome, with Chi² test.

By the way, you might perhaps find this useful in the future.

With kind regards

Karabiner

Hi Karabiner,

Thank you so much for your reply!
Yes, all dentists received both scenarios.

 

[priyac1987]

Hey,
if the original options ranged from 1-5 originally (1- strongly disagree, 2- disagree, 3- neutral, 4- agree, 5- strongly disagree) then it would have been okay to keep it that way and perform a regression test (H0: The patient's condition affects the surgeon's decision to perform a surgery). However if you now have a dichotomous question then chi square / cross-tabulation is better.
Option 3- if the questions presented to the surgeons were direct, then you may not need to perform any test. Simple frequency analysis would do.
Hope this helps.

 

[statskid2812]

Hey,
if the original options ranged from 1-5 originally (1- strongly disagree, 2- disagree, 3- neutral, 4- agree, 5- strongly disagree) then it would have been okay to keep it that way and perform a regression test (H0: The patient's condition affects the surgeon's decision to perform a surgery). However if you now have a dichotomous question then chi square / cross-tabulation is better.
Option 3- if the questions presented to the surgeons were direct, then you may not need to perform any test. Simple frequency analysis would do.
Hope this helps.

Hi priyac1987. Thank you for your reply. Let me try to clarify since the data is a little more complicated:

There is in total 20 cases, where the dentist had to choose in each case which surgery they would perform( there is 5 surgeries they can choose from, which Also includes the option "No surgery").

The first 10 cases is somewhat similar and have all one thing in common.
The other 10 cases is Also somewhat similar and have all one thing in common (which is the opposite of the first 10).

What I would like to know is if they perform more surgeries in the other 10 cases compared to the first 10 cases.

Hope this makes sense.

 

[priyac1987]

Thanks for explaining the problem. I think One-way ANOVA will be appropriate.

 

[Karabiner]

Thanks for explaining the problem. I think One-way ANOVA will be appropriate.

Could elaborate on this a bit? What is the interval scaled dependent variable, and what are the independent groups, if oneway anova is appropriate?
We have a dependent samples design with a binary outcome variable, as far as I can see, but maybe I am missing the point.

 

[statskid2812]

Thanks for explaining the problem. I think One-way ANOVA will be appropriate.

Thanks for the answer! The problem here is that I dont really have a dependent variable per se, I just want to look at the differences between the first 10 cases and the other 10 cases (whether or not there is performed more surgeries in the other 10 cases).

 

[Karabiner]

You clearly have a dependent variable, surgery yes/no.

 

[priyac1987]

So one variable would be coded value for the different type of surgeries and other variable would be this 10 case classification like head could be group and 1st 10 case can be stated as A while other 10 as B..Now dependent variable would be surgeries and factor of analysis would be group. Based on this analysis can be performed.

 

[Karabiner]

It is one and the same group of n=10 dentists. It is a repeated measures design.
Each dentist answered the "surgery yes/no" question two times, once for each of
two scenarios. Whether there is a statistically significant difference between
scenarios, can be analysed by using McNemar's test.