I'd be interested in reading your findings Alice Geraldine Akerman [log in to unmask]
Sent from my iPhone
> On 15 Nov 2013, at 00:14, PSYCH-POSTGRADS automatic digest system <[log in to unmask]> wrote:
>
> There are 5 messages totaling 1050 lines in this issue.
>
> Topics of the day:
>
> 1. Stats query! (2)
> 2. Assistance with piloting stimuli
> 3. Fwd: Stats query!/Private And Utmostly Secret
> 4. Development of a Pride Scale looking at Eating Attitudes - Volunteers
> needed
>
> ----------------------------------------------------------------------
>
> Date: Thu, 14 Nov 2013 13:27:52 -0000
> From: "Bennett, Alice [HMPS]" <[log in to unmask]>
> Subject: Re: Stats query!
>
> Hi Everyone,
>
> After looking into this logistic regression query further, I've got two
> more questions before I start my stats. I hope this is ok.
>
> I'm looking at what predicts personality disordered prisoners to drop
> out of prison based treatment. I've got data for 41 non-completers and
> 50 completers. I'm using a logistic regression.
>
>
> Question 1:
> Guidance that I've read suggests only using Stepwise for exploratory
> research (with a limited literature base). There is a developing
> literature base for this research question but the findings are very
> mixed, I'm wondering if this is as useful as no literature base (and
> therefore, pushing me to consider stepwise). Is this is a justified
> decision?
>
>
> Question 2:
> Because of the above mixed literature base, I have a lot of predictors
> available to me which I am aware is limiting in logistical regression.
> Would it be acceptable to run more than one logistic regression in this
> instance? Maybe this could inform what predictors to include in one
> final logistic regression and inform the research question better? Or is
> this mad/offensive to do?
>
>
> Any advice would be greatly appreciated.
>
>
> Best wishes,
>
>
>
> Alice Bennett
> Forensic Psychologist in Training
> Westgate Personality Disorder Treatment Service
> HMP Frankland
>
>
> Unclassified/Protect/Restricted
>
> P Please consider the environment before printing this e-mail or
> its attachments.
>
> This e-mail - and any attachment(s) - is intended only for the attention
> of the addressee(s). Its unauthorised use, disclosure, storage or
> copying is not permitted. If you are not the intended recipient, please
> destroy all copies and inform the sender by return e-mail. Internet
> e-mail is not a secure medium. Any reply to this message could be
> intercepted and read by someone else. Please bear that in mind when
> deciding whether to send material in response to this message by e-mail.
> This e-mail (whether you are the sender or the recipient) may be
> monitored, recorded and retained by the Ministry of Justice. E-mail
> monitoring / blocking software may be used, and e-mail content may be
> read at any time. You have a responsibility to ensure laws are not
> broken when composing or forwarding e-mails and their contents.
>
>
>
>
> ________________________________
>
> From: Research of postgraduate psychologists.
> [mailto:[log in to unmask]] On Behalf Of Jeremy Miles
> Sent: 31 October 2013 17:48
> To: [log in to unmask]
> Subject: Re: Stats query!
>
>
>
>
>
> On 31 October 2013 10:41, Bennett, Alice [HMPS]
> <[log in to unmask]> wrote:
>
>
>
>
> Hi Everyone,
>
> I'm hoping for some guidance/reassurance about which test to use
> on a study I'm working on.
>
> I want to know what differences there are between completers and
> non completers of a treatment service in order to inform management what
> best predicts those that do not complete treatment.
>
>
> I'd say it was the other way around. You want to determine what predicts
> who will complete, and who won't (otherwise your analysis is running
> backwards in time).
>
>
>
>
> My thoughts so far are.... my two unrelated IVs are: completers
> and noncompleters.
>
>
> I think you have one variable: completer, which can take on two values.
> Yes or no.
>
>
>
>
> The data I have includes the presence of various personality
> disorder traits/diagnoses (these are scored 0,1,2 which correspond to
> no, probable and definite diagnoses). My difficulty here is that as the
> numbers relate to someone having more evidence of a trait/personality
> disorder being relevant, that it is not separate categories as such. So
> I'm wondering if this is interval data (which would lead to parametric
> tests) or nominal data (chi squared test)
>
>
> Don't think in terms of tests being parametric. It's not a useful way to
> think about them.
>
> You're probably going to use logistic regression. Your predictor
> variables (0, 1, 2) can be treated as continuous, or categorical. If
> your sample size is huge, then treat them as categorical. If it's
> smaller, you're probably OK treating them as continuous.
>
>
> Separate to that, I also have interval data (such as scores on a
> psychopathy test/age) which I would also like to explore.
>
>
> You throw that in as a predictor. I'd suggest you do it hierarchically.
>
>
> Can anyone offer any advice? Or gin?
>
>
>
>
> Mmmm.... gin .....
>
> J
>
> This email was received from the INTERNET and scanned by the Government
> Secure Intranet anti-virus service supplied by Vodafone in partnership
> with Symantec. (CCTM Certificate Number 2009/09/0052.) In case of
> problems, please call your organisation's IT Helpdesk.
> Communications via the GSi may be automatically logged, monitored and/or
> recorded for legal purposes.
>
>
> The original of this email was scanned for viruses by the Government Secure Intranet virus scanning service supplied by Vodafone in partnership with Symantec. (CCTM Certificate Number 2009/09/0052.) On leaving the GSi this email was certified virus free.
> Communications via the GSi may be automatically logged, monitored and/or r
>
> ------------------------------
>
> Date: Thu, 14 Nov 2013 15:16:50 +0000
> From: Kelly Tate <[log in to unmask]>
> Subject: Assistance with piloting stimuli
>
> Dear all,
>
> Sorry to interrupt your day with a plea for help, but here goes...
>
> I require some help piloting stimuli for one of my experiments. The task is quite simple - you will be presented with various words and have to rate how positive, negative or neutral you think each word is. There are 30 words in total.
>
> Hopefully the task should take no longer than 10 minutes so if you have time to spare I will be very grateful! Just click on the below link to complete the task.
>
> Many, many thanks :)
>
> Kelly
>
> https://apps.mhs.manchester.ac.uk/surveys/TakeSurvey.aspx?PageNumber=1&SurveyID=8lLM79m3&Preview=true
>
>
> Kelly Tate
> PhD Researcher, University of Manchester
> Tel: +44 (0) 161 275 2589<tel:%2B44%20%280%29%20161%20275%202589>
> Sustainable Consumption Institute|188 Waterloo Place|Oxford Road|Manchester|M13 9PL|School of Psychological Sciences|Coupland 1 Building|Coupland Street|Oxford Road|Manchester|M13 9PL
> http://www.sci.manchester.ac.uk/people/postgraduate-students/kelly-tate
>
> ------------------------------
>
> Date: Thu, 14 Nov 2013 08:46:59 -0800
> From: Jeremy Miles <[log in to unmask]>
> Subject: Re: Stats query!
>
> On 14 November 2013 05:27, Bennett, Alice [HMPS] <
> [log in to unmask]> wrote:
>
>> Hi Everyone,
>>
>> After looking into this logistic regression query further, I've got two
>> more questions before I start my stats. I hope this is ok.
>>
>> I’m looking at what predicts personality disordered prisoners to drop out
>> of prison based treatment. I’ve got data for 41 non-completers and 50
>> completers. I'm using a logistic regression.
>>
>> Question 1:
>> Guidance that I've read suggests only using Stepwise for exploratory
>> research (with a limited literature base). There is a developing literature
>> base for this research question but the findings are very mixed, I’m
>> wondering if this is as useful as no literature base (and therefore,
>> pushing me to consider stepwise). Is this is a justified decision?
>
> Do you mean: When you're doing exploratory research, you should always use
> stepwise, or
>
> Stepwise is only OK when you're doing exploratory research.
>
> Stepwise anything is generally evil. It has a very high chance of giving
> you the wrong answer.
>
> There are much better things around than stepwise regression, but I don't
> think they've seeped into SPSS - things like boosted regression, lasso and
> LARS are preferred.
>
>
>
>> Question 2:
>> Because of the above mixed literature base, I have a lot of predictors
>> available to me which I am aware is limiting in logistical regression.
>> Would it be acceptable to run more than one logistic regression in this
>> instance? Maybe this could inform what predictors to include in one final
>> logistic regression and inform the research question better? Or is this
>> mad/offensive to do?
>
>
> Yes, I think that would be fine. You run a model with just demographics
> (age, sex, etc), then you run a model with demographics plus personality,
> then you run model with demographics plus (Oh, I don't know) crime type.
>
> The problem you're always going to have is that your sample size is small.
> This leads to two problems - you don't have much power. By my (very quick)
> calculation, if you have a simple dichotomous predictor and 20% of one
> group drop out you need 46% of the other group to drop out to have 80%
> power. That's a massive effect, it means that two and a half times as many
> people need to drop out in the second group for you to have a decent chance
> of detecting it.
>
> The second problem is worse, and that is that in logistic regression you
> run out of people very quickly, and that causes convergence problems and
> inappropriate solutions. In that previous example, I said that 20% dropped
> out - that's 8 people in one of your groups. Add a second dichotomous
> predictor, and if that's uncorrelated with the first, you're down to 4
> people in a cell, which is getting very low.
>
> Sorry to not be very positive.
>
> J
>
> ------------------------------
>
> Date: Thu, 14 Nov 2013 18:20:32 +0000
> From: Julie S Maclure <[log in to unmask]>
> Subject: Fwd: Stats query!/Private And Utmostly Secret
>
> Hi,
> I would suggest a graph plot of variable x v variable y, for your combination of predictors.
> I hope this helps.
> Julie S. MacLure MA( Honours) MEd MBPsS
>
>
> Begin forwarded message:
>
> From: Jeremy Miles <[log in to unmask]>
> Date: 14 November 2013 16:46:59 GMT
> To: [log in to unmask]
> Subject: Re: Stats query!
> Reply-To: Jeremy Miles <[log in to unmask]>
>
>
>
>
>
> On 14 November 2013 05:27, Bennett, Alice [HMPS] <[log in to unmask]> wrote:
> Hi Everyone,
>
> After looking into this logistic regression query further, I've got two more questions before I start my stats. I hope this is ok.
>
> I’m looking at what predicts personality disordered prisoners to drop out of prison based treatment. I’ve got data for 41 non-completers and 50 completers. I'm using a logistic regression.
>
> Question 1:
> Guidance that I've read suggests only using Stepwise for exploratory research (with a limited literature base). There is a developing literature base for this research question but the findings are very mixed, I’m wondering if this is as useful as no literature base (and therefore, pushing me to consider stepwise). Is this is a justified decision?
>
>
>
> Do you mean: When you're doing exploratory research, you should always use stepwise, or
>
> Stepwise is only OK when you're doing exploratory research.
>
> Stepwise anything is generally evil. It has a very high chance of giving you the wrong answer.
>
> There are much better things around than stepwise regression, but I don't think they've seeped into SPSS - things like boosted regression, lasso and LARS are preferred.
>
>
> Question 2:
> Because of the above mixed literature base, I have a lot of predictors available to me which I am aware is limiting in logistical regression. Would it be acceptable to run more than one logistic regression in this instance? Maybe this could inform what predictors to include in one final logistic regression and inform the research question better? Or is this mad/offensive to do?
>
>
> Yes, I think that would be fine. You run a model with just demographics (age, sex, etc), then you run a model with demographics plus personality, then you run model with demographics plus (Oh, I don't know) crime type.
>
> The problem you're always going to have is that your sample size is small. This leads to two problems - you don't have much power. By my (very quick) calculation, if you have a simple dichotomous predictor and 20% of one group drop out you need 46% of the other group to drop out to have 80% power. That's a massive effect, it means that two and a half times as many people need to drop out in the second group for you to have a decent chance of detecting it.
>
> The second problem is worse, and that is that in logistic regression you run out of people very quickly, and that causes convergence problems and inappropriate solutions. In that previous example, I said that 20% dropped out - that's 8 people in one of your groups. Add a second dichotomous predictor, and if that's uncorrelated with the first, you're down to 4 people in a cell, which is getting very low.
>
> Sorry to not be very positive
>
> ------------------------------
>
> Date: Thu, 14 Nov 2013 18:45:36 +0000
> From: Cintia Faija <[log in to unmask]>
> Subject: Development of a Pride Scale looking at Eating Attitudes - Volunteers needed
>
> Dear colleagues,
>
> My name is Cintia Faija, I am a full time PhD student in Psychology at The University of Manchester and my supervisory team is comprised of: Dr. John Fox, Dr. Patricia Gooding and Dr. Stephanie Tierney.
>
> Are you interested in research looking at attitudes and emotions towards eating? If yes, I would like to to invite you to take part in an anonymous on-line survey study, which take no longer than 30 minutes to complete and involves questions on eating attitudes, self-esteem and pride feelings. The aim of the study is to test the psychometric properties of a recently developed measure that assesses pride towards body weight, body shape and eating attitudes.
>
> The study has been awarded ethical approval by the Faculty of Medical and Human Sciences, The University of Manchester (Approval received: 05/07/2013, Ethics Number: 13047).
>
> If you are willing to participate please follow this link:
> https://apps.mhs.manchester.ac.uk/surveys//TakeSurvey.aspx?SurveyID=92LM7n3K<https://apps.mhs.manchester.ac.uk/surveys/TakeSurvey.aspx?SurveyID=92LM7n3K>
>
> If you would like any further information, please do not hesitate to contact me at:
> [log in to unmask]<mailto:[log in to unmask]>
>
> I would greatly appreciate your participation!!
>
> Many thanks,
>
> Cintia Faija.-
>
> Cintia Faija
> PhD Student
> School of Psychological Sciences
> Room H22, Ground Floor, Coupland Building 1
> University of Manchester
> Oxford Road
> Manchester
> M13 9PL
>
> ------------------------------
>
> End of PSYCH-POSTGRADS Digest - 13 Nov 2013 to 14 Nov 2013 (#2013-256)
> **********************************************************************
|