i did already upload a document! i did already upload a document! i did already upload a document!

Download I did already upload a document! I did already upload a document! I did already upload a document!

If you can't read please download the document

Upload: jaquelinessrim

Post on 04-Sep-2015

5 views

Category:

Documents


3 download

DESCRIPTION

I did already upload a document! I did already upload a document! I did already upload a document!

TRANSCRIPT

Content validation: working with a subject-matter expert, history, or some other authoritative content, use human judgment to demonstrate that your survey instrument appears to measure what it's supposed to measure. You can use a focus-group/interview type of process to collect these human judgments. Consider quantifying rater agreement by using multiple human judges (subject-matter experts, SMEs), and/or consider revising or deleting items that produce agreement outliers.Internal validation: explore the quantitative characteristics of your survey items and sub-scales, and use consistent and fair statistical significance tests to demonstrate that if phenonemon A produces a distinct response pattern in some portion of your instrument, then other portions of your instrument that purport to measure phenomenon A produce the same result. This is related to inter-item reliability. Conventional statistical tests might include cronbach's alpha estimates on the sub-scales in your instrument. Values greater than .60 may be worth working through; over .80, you are getting closer; into the .90s, you are finally getting some defensible evidence of internal validity. If your instrument is used for high-stakes decisions (like personnel selection), you will want as high levels of reliability as you can get (like high .90s).Factor analysis can be used across your sub-scales (if your instrument collects information on multiple phenomenon) to provide more compelling evidence of the construct validity of your instrument. A confirmatory factor analysis that results in your predicted factor loadings having an acceptable model fit can give you a lot of confidence about your instrument; this is pretty rigorous evidence if you can produce it.Construct validity is generally an important concept and is related to the other validity evidences you mentioned. Part of a strong case for construct validity is in demonstrated that your instrument is not just measuring some other similar phenomenon.External validity will come with time as you implement your survey in different samples. A good expert review from your content validation exercise should help you build a survey that captures the most important and typical indicators of your phenomenon; this should help with external validity. However, the search for external validity will ultimately depend on how widely and for what purposes you intend to use your measure. As far as a i know, there is not a clear standard for this type of evidence (except maybe showing that your factor structure is consistent across samples), and it may be a little unreasonable to ask any single investigator to definitively establish the external validity of any instrument. This should be a field-level effort after you have had a chance to use your instrument in different contexts. Perhaps meta-analysis is a good systematic method for establishing evidence of external validity.