Attribute Recall v Attribute Importance
Locked
 
|
Seems as though the easiest question leaves a blank, so a nudge please :)
Have a series of attributes whereby the response is "Yes, recall seeing or hearing about it in the last 30 days" versus "No, don't recall seeing or hearing about it in the last 30 days" Next for the same set of attributes: "Yes, is important to me" and "No, is not important to me" I am looking to understand if a relationship exists between recall and importance. I also know the responses are poor but had no choice in that. Have looked at simple correlations but I just know there is something else much more viable, IF there is? Tks ----- Will Statistical Services ============ [hidden email] http://home.earthlink.net/~z_statman/ ============ -- Sent from: http://spssx-discussion.1045642.n5.nabble.com/ ===================== 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
Will
Statistical Services ============ info.statman@earthlink.net http://home.earthlink.net/~z_statman/ ============ |
|
My first question is how do researchers analyze this type of response data? There are a variety of ways to analyze the situation you describe but you describe only one. Let me describe two possibilities but may not be acceptable to researchers in the relevant area(s), First, instead of analyzing pairs of responses (i.e., recall vs importance for each question), analyze all responses for each subject/participant. For example, if you restructure the data so that each subject has two columns containing the recall-importance pairs, you can calculate the Pearson r for each subject (assuming you use 0,1 coding where 1= recall-yes, importance-yes and 0= recall-no, importance-no; the Pearson r is actually a phi coefficient -- one could also calculate the odds ratio instead or in addition). This analysis provides a Pearson r for each person and indicates if the pattern of responses are related. If positive, then recall and importance appear to operate together. If r -0.00, then recall and importance operate independently. If negative, when recall is high, importance is low and vice versa. Then calculate the median Pearson r (phi) across subjects and.or divide subjects into groups on the basis of the sign of the correlation. Relevant theory should be used to interpret the results though the above interpretations should be valid to a first approximation. There are two justifications for the above analysis: (1) Stephenson's Q methogology was developed back in the middle of the 20th century to systematically study relationships between subjects. Use Google scholar to find relevant publications. (2) Saul Sternberg's "memory scanning exoeriments" had subjects determine whether a target (e.g., the letter "H") was present in a list of items (lists varied from one item to five items). After several hundred trials (each trial had a different list and list length was randomized across trials) the responses (i.e., Reaction Time or amount of time it took to make a response, typically in thousands of a second or milliseconds [ms]). Results can be analyzed on a subject by subject basis and/or using group data. Sternberg found that subjects engaged in exhaustive serial processing even when the target was not on the list (i.e., subjects did not engage in "self-terminating search" after detecting the item on the list; if the item was the first element in the list, people still scanned the entire list). Sternberg's reults indicate that people take 30 ms (point estimate) to scan a list in memory, an extremely fast memory process. However, theory is important because subsequent researchers have argued that (a) the search is not serial but parallel with all items being examed at the same time (see Theios work) and (b) one has to beleve in a traditional memory model where short-term memory (STM) has properties different from long-term memory (LTM); today most researchers accept some version of Baddeley's working memory model (WM) or assert that STM is actually the temporary activation of LTM representations. The preceding should give some sense of rhe relevance of theory in making an interpretation of results. Second, another way to analyze your data is to get the sum or mean for the recall items (coded 0,1) and the sum or mean of the importance items (coded 0,1) and then forming a ratio for each subject like the following: ratio = (sum or mean recall)/(sum or mean importance) A ratio > 1.00 indicates that recall considerations appear to be more relevant than importance, a ratio = 1.00 indicates that recall and importance appear to play similar roles, and a ration < 1.00 indicates rhat importance is, well, more important than recall. The above anallysis is coarse in that it relies on a summary measure across two different 30 item vectors. One could do a fine grain analysis by getting the ratio for each item and getting a summary measure (e.g., a median) for each subject. It is possible that neither of the above analyses are used by researchers in this area but you should know what is commonly done. So, if you don't know what analysis is commonly done, list to Hamlet: "Get thee to Google Scholar. Or PsycInfo if you have access." :-) -Mike Palij New York University On Sun, Jan 27, 2019 at 11:08 AM zstatman <[hidden email]> wrote: Seems as though the easiest question leaves a blank, so a nudge please :) |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |