Pat Bazeley

To get an overview of each of your themes as well as being able to see if they are a source of satisfaction or dissatisfaction, make the two satisfaction nodes children in a separate tree, and organise your 'themes' in other trees (preferably according to the kinds of things that are raised, e.g. separating interpersonal issues from infrastructure issues). If you then code each passage for the issue raised and also for whether it was seen as satisfactory or not (i.e., at least two codes), then you can run a matrix to see what the pattern of responses is. To add in additional factors like your groups, use the 'In > selected items' option in the matrix dialogue. You cold also save the results of the first matrix as a tree node, and use the lower level children in that in a matrix query against your groups to get everything at once (might take a bit of computing time!) Alternatively, to just compare what was positive for all your subgroups - run the groups by themes, In > text coded at positive, etc. Save the query, and repeat for negative. Pat

That would be even better if you could have the node tree open at different places on each side! Pat

No. If a code is deleted the order would remain the same, but not if a new one is inserted. Most people don't use a numbering system for nodes. As a default, NVivo sorts them alphabetically, but there is a custom sort button on the toolbars (looks like 3 dots arranged vertically) and if you click on it you can then move nodes up or down. Next time you open, it will appear alpha order, but a single click on the icon will revert it to your custom order. Pat

Hi Terry, If the sources are identical, then they should have merged into a single source. Perhaps you gave them different names to reflect who coded, or perhaps each person couldn't resist the temptation to edit the source in some way? For the coding comparison query, they need to be a single source - NVivo can detect who did what coding on it, and show the level of agreement both with Kappa and with coding stripes for each coder (double click on a line in the coding query results). Pat

No. Case counts in an attribute by node matrix, as a proportion of the total number with the relevant attribute value, has been requested. Pat

No. The fastest way to add the same code to further passages of text is, probably, to highlight the next passage and click on the tick next to the recently used node slot in the coding toolbar (the last used node will be showing). Pat

Here's a couple of other possibilities: a. In version 8 you can display attribute values as coding stripes. b. Copy the saved matrix to the trees area, then select all the nodes at the third level down (= cells) and export them (with hierarchical names) in one go. Create a Word document and insert all the exported files at once. Then run a Replace which finds 'Name' and replaces it with Name in a heading style (use the More button in the Replace dialogue), then use the outline view to reorder the headings (with their associated text) to what you want. c. Copy and paste the matrix table into Word, and add a brief summary in each cell of the material that is coded there (this is a good analytical strategy). Cheers, Pat

Hi Terry, Why not use a matrix coding query? Then the combinations would be in separate cells for each attribute (listed in the order of the attributes in the selection box (which can be modified - assuming version 8). Text within results is ordered firstly by folder, then by source name. Pat

Options for coding to cases in NV7 or NV8 Code individual sources as cases on import. (Tidy up your document names before importing!) Or, after importing your sources: If your Cases are represented by single sources: Select the sources representing Cases in the List View, right-click, and choose Create As > Create Cases. (This will make a case node for each source.) If your Cases involve multiple whole documents: Select the sources representing a single Case in the List View, right-click, and choose Code Sources > At New Node. Select Cases as the location, and then name the new node. If several cases are combined within a source document (e.g., from a focus group): Prepare the document by using consistently applied heading (paragraph) styles to identify each case (cf. data preparation notes on my web site). Auto code the document. Pat

The best solution to seeing who says something you've coded is to turn on the coding density bar - because the speaker has been made into a case, the case node will be listed there. Pat

Not if you want to be able to auto code them - you need some form of heading in there. What you might do is reduce the headings to something much smaller (and a standard size, e.g. Q1 etc) rather than having a whole question there. You might also want to split up your source document into smaller sections (it won't cause any problem to do so as long as you don't split in the middle of a case) as NVivo can become very slow when coding a very long document (it has to count characters to reference the location of the coded text at the node). Pat

Hi Betty, Users are one of the few categories of project item that you can't choose from when setting up rows or columns in a matrix (though you can limit the query to one user). You could either use codes across whole patient cases for hard and easy, or you might prefer,if they are included sequentially within a transcript (i.e. they are not all mixed up), to insert a unique heading for each into the transcripts and then use those headings to create patient cases by auto coding them (check the help files under auto code). The problem with that is likely to be that, if you have different people working on the same transcripts in different copies of the project and then want to merge them, they won't merge because they will have become different - so you'd really need to leave doing that until it was all coded and already merged into in a single copy. What the autocoding will do is effectively 'cut up' the transcripts for you, without making separate documents of them. What you'd need to remember to do then is to ask for the count of cases rather than sources when you are running matrices (but if you still have the whole sources as cases that also will create a counting issue as text will be in two cases). Pat

HI Bruce, If the second person sets up a project they can import your project without coding (under File > Import Project - check the options out. Pat

Hi Keith, You will probably find the book I've written on 'Qualitative data analysis with NVivo' useful (Sage 2007). It was written when NVivo 7 came out, so some of the instructions have changed but the general tenor of the explanations about what you might do using the software, and why you are doing it (with reference mainly to qual research in general, but also occasionally with reference to specific methodological approaches). (Updates to the instructions for version 8 are available on my website.) Also on my website, on the mixed methods resources page, you will find references to (and for some, access to the text of) a number of articles I've written on analysing and/or using computers to analyse mixed methods data, including from structured questionnaires. There is also an article on an approach to qual analysis in general (under working papers on the qual resources page) which people have found helpful. Pat

There's no way a text query for compliance would find material at the node if it didn't include the word. What you might be seeing is context around the finds you specified - NVivo shows 5 words either side, in a lighter grey text, by default, as context for your finds. The other thing to watch, when you save your finds using the options, is what context has been specified for saving. There is a way that you can exclude a node from a text query which is useful if you want to check the thoroughness of your coding: Use a compound query, putting the text query criteria into the first sub-query slot, choosing AND NOT for the operator in the middle, and then change the bottom sub-query to a coding query, and simply choose the node (e.g. for compliance) in there and you will get finds that are not already coded - saves a lot of checking time on the results. Pat