What to do with this situation?

[Maguin, Eugene]

Somebody wrote a survey and intended to use a standard 5 point likert type response format (Strong D, Some D, Neutral, Some A, Strong A) but by mistake used this format (Strong D, Some D, Neutral, Some A, A, Strong A). So two points below Neutral and three points above. I’m flummoxed about advice to give other than to say to analyze the data using different response treatments and discuss the treatment effect differences in footnotes. Can you help me give better advice?

Gene Maguin

===================== 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

Hi, Gene,

not sure about the background but you can think about merging answers into categories

Disagree - Neutral - Agree

or even

Strong D, Some D, Neutral, Agree

Am Mittwoch, 14. November 2018, 20:35:59 MEZ hat Maguin, Eugene <[hidden email]> Folgendes geschrieben:

Somebody wrote a survey and intended to use a standard 5 point likert type response format (Strong D, Some D, Neutral, Some A, Strong A) but by mistake used this format (Strong D, Some D, Neutral, Some A, A, Strong A). So two points below Neutral and three points above. I’m flummoxed about advice to give other than to say to analyze the data using different response treatments and discuss the treatment effect differences in footnotes. Can you help me give better advice?

Gene Maguin

===================== 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

===================== 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

[Rich Ulrich]

If they can go the sophisticated way, they may use Item Response theory. That will

effectively re-score all the intervals - so the initial labels don't matter. I've never used IRT.

My original concern about it was that it might require large samples, but examples seemed

robust with samples of 50, or maybe 30. It's been quite a while since I read them.

Googling shows me http://www-01.ibm.com/support/docview.wss?uid=swg21488442

which describes several R plug-ins that can be used from SPSS.

Is it possible that the original categories were chosen conscientiously, rather than by

accident?  - If they can argue that those represent a reasonable estimate of "equal

intervals", that might answer complaints that the scale is not your ordinary likert-type

intervals.

--

Rich Ulrich

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of Maguin, Eugene <[hidden email]>
Sent: Wednesday, November 14, 2018 2:35 PM
To: [hidden email]
Subject: What to do with this situation?

 

Somebody wrote a survey and intended to use a standard 5 point likert type response format (Strong D, Some D, Neutral, Some A, Strong A) but by mistake used this format (Strong D, Some D, Neutral, Some A, A, Strong A). So two points below Neutral and three points above. I’m flummoxed about advice to give other than to say to analyze the data using different response treatments and discuss the treatment effect differences in footnotes. Can you help me give better advice?

Gene Maguin

===================== 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

===================== 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

[Kirill Orlov]

Have of A could be binned into SomeA and half into StrongA because it is ordinal scale.

An alternative could be to merge it with one of the two adjacent points completely. One might use Categorical PCA to obtain the quatification that will suggest which is better.

If there is a predetermined dependent variable (while that/these ordinal ones are seen as predictors - then one might likewise use Classification tree CHAID to see if merging A with SomeA or StrongA is better.

14.11.2018 22:35, Maguin, Eugene пишет:

Somebody wrote a survey and intended to use a standard 5 point likert type response format (Strong D, Some D, Neutral, Some A, Strong A) but by mistake used this format (Strong D, Some D, Neutral, Some A, A, Strong A). So two points below Neutral and three points above. I’m flummoxed about advice to give other than to say to analyze the data using different response treatments and discuss the treatment effect differences in footnotes. Can you help me give better advice?

Gene Maguin

===================== 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

===================== 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