AQMeN - Applied Quantitative Methods Network

Hi Helen

It is hard to give a general answer as it depends on the content of your PhD, eg what methods were used, were there sampling issues. Could you give a bit more detail?

Thanks
Helen

You need to be able to show a degree of mastery of methods and statistics. I advise asking for a mock viva. you show mastery by being able to answer questions...

Hello,

Thanks for your replies. I a longitudinal study in schools across New Zealand that used two questionnaires completed approximately 12 months apart. I used several scales that were already available from other published studies and put them into my questionnaire, as well as putting in some questions of my own.

The questionnaire items used 5 and 7-point likert scales. The theoretical framework underpinning my research was the theory of planned behaviour and extending it to incorporate a measure of past behaviour.

I ran the following tests on the data:
*Factor analysis (using PCA and PFA)
*Cronbach's Alphas
*Independent samples t -tests (to look for significant gender differences etc.)
*One way ANOVA's (to look at sig differences according to location of adolescents eg. city, rural, suburban areas)
*Correlations (to explore relationships between variables)
*Regressions - I used forward regressions so that only those items that significantly added to the model were shown.

Because I was looking at the addition of past behaviour to the TPB model I probably should have used confirmatory factor analysis (i.e. structural equation modelling) but the software was unavailable at the university. Also, I was told that regressions were fine....to be honest, stats is my achilles tendon, hence why I joined this network!!!

Helen

Hi Helen

Probably the best preparation would be to understand the techniques you have used and why in as much detail as possible, plus have some knowledge of those you might have used but didn't. Finding out about the level of statistical knowledge of your examiners might also be helpful! Plus you could request a consultancy session with us - if this is of interest just complete the form at http://aqmen.ac.uk/Consultancy.

Helen

There are free SEM programs.... but it's not helpful to suggest what you should have done.
It'll depend on the questions your examiners will ask! the trouble with the range of techniques you've used is that each has a wide range of issues when applied to items and scales.. And you've got to know enough about what the issues are be able to answer questions that might (or might not) be forthcoming. Researching the expertise of your examiners is thus worthwhile.
If I was an examiner I'd be asking "How would you have used a SEM package?" but that would be a bit cruel.. (expecting somethign about measurement model, Structural models for the regression, and even group differences).
Perhaps your best strategy would be to divert the questions back to the content area, theory etc... If your results are clear and unequivocal it doesn't make much difference which techniques you use. It's the interpretation of the results that matter.

As well as Helen's and Steve's valuable suggestions, I'd say that you should show that you have tried out lots of different angles on the data. I always think of the statistics that appear in any published work (including a doctoral thesis) as being like the tip of an iceberg: far more analysis gets done than is ever reported, but without that background you can't be sure that your analysis is robust.

Hi Helen
I too will be facing my viva soon and my thesis is both quants and qualit. I have tried to think about how to defend my stats and I have been heartened by the replies you have had. I really think this is sound advice. I spent far too much time doing lots of stats, creation of variables, splitting analyses and so on and now I realise it is not even being used as such. so, this gives me some scope to prove I did do the exploratory stuff by trying out different tests, tried things, saw mistakes, better ways and then reaching a decision. For e.g. deciding whether I needed to transform the data took me ages - I just could'nt commit to a quick decision on this - i.e. how normal is normal? even although my supervisor was there to assist. With hindsight, I appreciate why he let me do it for myself - as at the end of the day, it's me that will be examined on how well I know my work and why I did things the way I did them.
best wishes
Carol

Hi Helen,
i have serious concerns that your use of stepwise regression will be thrown out - as noted in another post i know no-one who would seriously present a stepwise regression in an published paper - it would not get past peer review for a start.