analysis of a survey
Locked
 
|
Dear all, As mentioned earlier, I
would like to ask your help in a delicate problem. We have done a
survey for physicians. Here, I would like to describe it more precisely.- n=495 physicians participated - first, we focused on their experience with four different diagnostic tools. For each they had a single-choice question (no experience / less than 1 year / between 1 and 3 years / more than 3 years) - second, we presented cases (5 in total, the same cases for all the participants. Cases are considered to be identical in terms of difficulty or type of disease, etc.). Here they were asked to choose one out of the 5 potentially useful diagnostic tools. - the aim of the study is to evaluate whether there is any correlation between the level of experience with different diagnostic tools and the decision patterns. The original idea was to do the analysis on case-basis. - note, that not all the participants reviewed all the five cases, a few left earlier the questionnaire. So far I understood my problem I should perform a multivariate, mixed, multinomial, logistic regression model, since Is it correct? Could you explain me stepwise how can it be performed?Thx++++ |
|
|
My response here strikes me as Off Topic for "SPSS usage"
since I am offering more of a review of the considerations for analysis. You get dumb analyses if you don't consider what sort of conclusions *your* data can support, despite the structure apparent in the design. Your survey is limited to five "cases" to be considered for recommending diagnostic procedures. I'm afraid that I can too- easily imagine that there are drastic limits on what you can conclude, if the cases were not well chosen, and the question was not carefully asked. (Was there pilot data?) I'm sure that I would look first at the individual cases to make sure that I understand what is going on, before I considered combining all answers to look at Repeated Measures or other multivariate approach for DVs. For instance: Is there a "right answer"? Is there a single answer than the best physicians will agree on? (Should you look for that?) Having a right answer or best answer is more likely to be true for a case if you asked, 1) "Which diagnostic will give the best answer?" - implicitly or explicitly ignoring cost to the patient in money, time, pain, side-effects, etc.; instead of asking 2) "Which diagnostic would you use first for this patient?" If there is a right answer, that changes the focus of the interpretation. If almost everyone agrees on the right answer, that puts a real crimp into trying to look at variation. So: Is there any variance in the answers, for each of the cases? Are all of the cases still worth considering, or do you need to drop one or more from further consideration, after describing the uniform responses? Further, is there any variance in how Experienced they are? On reading the scale, Experience is a continuous variable. However, I can easily imagine that a random 400 of 495 doctors have 3+ years experience with X-rays. Or 300 of them might have 3 years experience with everything. Your power for looking at differences in experience will be drastically reduced if there is few people with low experience. - Do you have anything else (Age? specialty?) to incorporate a measure of experience? - Collapsing a variable to a dichotomy is very often a bad idea, but if there are hardly any with less than 3 years, that might allow an easier and fairer description. -- Rich Ulrich ________________________________ > Date: Thu, 19 Dec 2013 07:26:41 +0100 > From: [hidden email] > Subject: analysis of a survey > To: [hidden email] > > Dear all, > > As mentioned earlier, I would like to ask your help in a delicate > problem. We have done a survey for physicians. Here, I would like to > describe it more precisely. > > - n=495 physicians participated > > - first, we focused on their experience with four different diagnostic > tools. For each they had a single-choice question (no experience / less > than 1 year / between 1 and 3 years / more than 3 years) > > - second, we presented cases (5 in total, the same cases for all the > participants. Cases are considered to be identical in terms of > difficulty or type of disease, etc.). Here they were asked to choose > one out of the 5 potentially useful diagnostic tools. > > - the aim of the study is to evaluate whether there is any correlation > between the level of experience with different diagnostic tools and the > decision patterns. The original idea was to do the analysis on > case-basis. > > - note, that not all the participants reviewed all the five cases, a > few left earlier the questionnaire. > > So far I understood my problem I should perform a multivariate, mixed, > multinomial, logistic regression model, since > - outcome is multinomial > - there are repeated measures from the same participants > - want to evaluate the impact of multiple conditions > > Is it correct? > > Could you explain me stepwise how can it be performed? > > Thx++++ ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
|
In reply to this post by Dr. Gábor Tóth
How many physicians are there in your sample, and how were the data collected? How delicate is “delicate” Have you actually looked at the basic distributions in frequency and contingency tables yet? If not, I suggest you do that first. You can tell a lot just from %% and charts without getting into multivariate modelling: that can come later. John F Hall (Mr) [Retired academic survey researcher] Email: [hidden email] Website: www.surveyresearch.weebly.com SPSS start page: : : www.surveyresearch.weebly.com/1-survey-analysis-workshop From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Dr. Gábor Tóth Dear all, I would like to ask your help in a delicate problem. We have done a survey for physicians, where they were asked to make a decision (choose one out of five options) for five independent clinical cases. We aim to analyze their decisions according to their experience in different fields medicine which they defined by putting themselves into one of four categories for each. - outcome is multinomial - there are repeated measures from the same participants - want to evaluate the impact of multiple conditions Could you explain me stepwise how can it be performed? |
| Free forum by Nabble | Edit this page |