Re: Personality questionnaires

On 22 Sep, 21:46, Lance <LanceG...@xxxxxxxxx> wrote:
On Sep 22, 12:59 am, Dave Smith <da...@xxxxxxxxxxxxxxxxxxxxxx> wrote:





On 21 Sep, 10:07, Lance <LanceG...@xxxxxxxxx> wrote:

On Sep 21, 1:06 am, "Philip" <pp...@xxxxxxxxxxxxx> wrote:

"Lance" <LanceG...@xxxxxxxxx> wrote in message

news:1190323156.062474.110930@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

On Sep 20, 11:08 pm, Dave Smith <da...@xxxxxxxxxxxxxxxxxxxxxx> wrote:
On 20 Sep, 13:36, Lance <LanceG...@xxxxxxxxx> wrote:

On Sep 19, 12:01 pm, "Peter H.M.Brooks" <pe...@xxxxxxxxxx> wrote:

Lance wrote:
On Sep 18, 11:02 am, Dave Smith <da...@xxxxxxxxxxxxxxxxxxxxxx>
wrote:
This article -- a trailer for a book -- may be of interest. It
provides a non-technical (and uncritical?) account of the 'big
five'
personality dimensions' and how they relate to other variables.

http://news.independent.co.uk/sci_tech/article2971114.ece

I think it is good article. I'm not sure it will be to PeterB's
taste,
but it does set out what is a growing consensus amongst
psychologists.

I don't think that I've got anything against it. I am pleased that
personality factors are being teased out in a repeatable manner. As
you
know, I would like the enneagram to be investigated more, and this is
happening, there is a repeatable inventory and mappings to other
personality inventories have been found to be fairly reliable:

http://www.enneagraminstitute.com/research.asp

I think there is a fundamental difference between types and
dimensions. The big five, as Dr Nettle says, are really dimensions,
and, a priori, anyone could be located anywhere on the five
dimensions.

It is quite possible, given that the five dimensions really exist,
that particular positions in five dimensional space occur more
frequently that would be expected purely by chance. In that case we
can define a type as a group of people who fall in a particular
location of five dimensional space as defined by the Big Five more
frequently than chance expectation. if you simplified the five
dimensions to two opposite categories (neurotic, not neurotic,
agreeable, not agreeable), then a type would be a combination of
categories that occurred more frequently than one would expect on the
basis of the marginals.

Give the above analysis, it sems to me that a system of types may be
completely compatible with the Big Five, and may some understanding to
the details of how they work in real people.

But I expect scores on the five dimensions have bell-shaped, uni-modal
distributions.

Dave

The interaction between the five dimensions may not be normally
distributed. In fact the interaction of the five may well create types
(and anti-types).

That seems to be inconsistent with Nettle's assertion "Your score on one
dimension is independent of your scores on all the others". Is he wrong?- Hide quoted text -

- Show quoted text -

Nettle is not wrong. The independence of the dimensions does not mean
that typologies can't be found - in fact, if you are to find them, you
need some assumptions on which to base estimates of how likely
particular combinations may be, and that requires assuming
independence of the dimensions. Have a look at textbooks by people
like Alexander von Eye for details of how this can be done.

What do you make of this very simple example? Say, we have four types
of person positioned on three dimensions. Type 1 score low on all
three dimensions. Type 2 score low on the first dimension, high on the
second dimension and high on the third. Type 3 score high on the first
dimension, low on the second and high on the third. And Type 4 score
high on the first and second dimension and low on the third. With
rows representing the four types, columns representing the three
dimensions and 'L' and 'H' representing low and high scores we can
present our example schematically like this:

LLL
LHH
HLH
HHL

It would seem that the three dimensions could be orthogonal to each
other, since scores on each pair of dimensions show the four
combinations LL, LH, HL and HH. Yet the four types are to be found in
only half the three-dimensional space.

If I'm right, this example illustrates how it is possible to have
completely discrete types defined by orthogonal dimensions. However,
such an outcome seems unlikely, since if the types don't 'balance'
the orthogonal structure breaks down. In the example, if type 1 are
particularly numerous there will be many LL combinations and the
dimensions will thus tend to correlate positively with each other --
and so on.

I did quite a lot of research with personality scales many years ago,
and I found scores on the scales typically had a bell-shaped uni-modal
distribution. Correlations between scales are somewhat arbitrary,
since they reflect the choice of items. For instance, if an
extraversion scale mostly contains questions about sociability, then
scores on it are likely to correlate negatively with scores on a
neuroticism scale. (This is probably because 'highly anxious'
individuals avoid social situations.) If some of the items about
sociability are replaced with items about impulsiveness, then the
negative correlation between extraversion and neuroticism can be
reduced more or less to zero.

It is too late at night for me to try to think through all the details
of your example. I will just say that perhaps your example is too
simple. Configural frequency analysis tends to be applied to very
large samples and to to seek types that arise in such samples. Given
that there are nine eneargram types and that there are millioins - no
billions - of people who fall into the five personality dimensions it
would seem really surprising to me if not types and no anti-types were
to be found. And yes normality in the univariate case doesn't
necessarily imply normality in the multivariate case, and we are
clearly looking at multivariate data here.

Anyway, I wasn't trying to challenge you - i was just thinking at the
keyboard, wondering if such a possibility could not arise? I haven't
done any data analysis on the topic. It was just a thought.

I think personality typologies are usually held to apply to a
substantial proportion of the general population, or at least to a
substantial proportion of a particular population such as 'juvenile
delinquents' -- it is not a matter of showing that there are several
people who form a distinct type, it is a matter of showing that an
appreciable percentage of a particular population form a distinct
type. I don't think the absence of such distinct types could be
proven, but it seems to me that it is up to those personality
theorists who advocate a particular typological approach to
demonstrate its reliability and validity.

.

Relevant Pages

  • Re: Personality questionnaires
    ... personality dimensions' and how they relate to other variables. ... personality factors are being teased out in a repeatable manner. ... dimensions will thus tend to correlate positively with each other -- ... I did quite a lot of research with personality scales many years ago, ...
    (uk.philosophy.humanism)
  • Re: An experimental design to distinguish personality types
    ... from the Enneagram personality type inventory we get a point ... If there is anything to the personality types in the Enneagram (and this ... distance from that cog and a standard deviation of those mean distances. ... the dimensions used are not satisfactory in some way (e.g. they are ...
    (uk.philosophy.humanism)
  • Re: Personality questionnaires
    ... personality dimensions' and how they relate to other variables. ... personality factors are being teased out in a repeatable manner. ... dimensions will thus tend to correlate positively with each other -- ... Correlations between scales are somewhat arbitrary, ...
    (uk.philosophy.humanism)
  • Re: Personality questionnaires
    ... personality dimensions' and how they relate to other variables. ... personality factors are being teased out in a repeatable manner. ... Correlations between scales are somewhat arbitrary, ... negative correlation between extraversion and neuroticism can be ...
    (uk.philosophy.humanism)
  • Re: Personality questionnaires
    ... personality dimensions' and how they relate to other variables. ... personality factors are being teased out in a repeatable manner. ... dimensions will thus tend to correlate positively with each other -- ... I did quite a lot of research with personality scales many years ago, ...
    (uk.philosophy.humanism)