Hi all,
My survey elicits people's use of three different languages in different contexts. For each context, participants are supposed to indicate how frequently ("Never," "Seldom," "Sometimes," "Often," and "Always") they speak each language (i.e., Putonghua, English, and Cantonese). However, when inputting the data, I have identified the following problems in the survey:
(1) In the survey, the instruction indicates that participants should tick all the three languages in each context regarding the use frequency, but in reality, some of them just ticked one or two languages and left the rest blank. Should I discard this kind of answer? Or Can I just treat it as a missing value?
(2) In one context, some participants chose "Always" for more than one language, which seems contradictory with the fact that people cannot "'always" speak two or more than two languages in one context. (BTW, code-mixing is possible, but I don't take this behavior into account.)
(3) I intend to use this set of data to answer the following two questions: How frequently are Putonghua, English, and Cantonese used in different contexts? Is there any significant different among the three languages in one context? I attempt to calculate the mean values for each language in each context and then do the multiple comparisons. Does it make sense?
Thank you all, and looking forward to your suggestions.
Cheers,
Richard
Dear researcher,
Regarding your first question, the usual procedure is to consider these questions with missing answers. Nonetheless, if the part that was not-answered is not the same in all cases, you may assume that the answers were “never”, which can be confirmed via contacting any of these respondents or by the absence of any part being answered only by “never”.
Regarding the second question, answering by “always” for two or more different languages can be explained that the respondent was considering different contexts for each language, for example different teachers or classes. Therefore, you may accept these answers with the absence of any conflict.
Good luck
Dear Richard--May be the participants who have not responded to the frequency of using the languages but have indicated only the languages they speak have to be considered as `attrition', as both your aims seem to be linked to the measure of frequency of the languages used (and hence a likert and not a dichotomous endorsement). Treating as missing values even with multiple imputation could lead to misleading findings. Cheers.
- Badges
- Science topic
- Similar topics
- Linguistics
- Philology
- Language