OpenRepGrid (version 0.1.9)

distanceNormalized: Standardized inter-element distances' (power transformed Hartmann distances).

Description

Calculate power-transformed Hartmann distances.

Usage

distanceNormalized(x, reps = 1000, prob = NULL, progress = TRUE,
  distributions = TRUE)

Arguments

x

repgrid object.

reps

Number of random grids to generate to produce sample distribution for Hartmann distances (default is 1000). Note that a lot of samples may take a while to calculate.

prob

The probability of each rating value to occur. If NULL (default) the distribution is uniform. The number of values must match the length of the rating scale.

progress

Whether to show a progress bar during simulation (default is TRUE) (for method="simulate"). May be useful when the distribution is estimated on the basis of many quasis.

distributions

Wether to additionally return the values of the simulated distributions (Slater etc.) The default is FALSE as it will quickly boost the object size.

Value

• A matrix containing the standardized distances. Further data is contained in the object's attributes:
• "arguments"
  A list of several parameters including mean and sd of Slater distribution.
• "quantiles"
  Quantiles for Slater, Hartmann and power transformed distance distribitions.
• "distributions"
  List with values of the simulated distributions, if distributions=TRUE.

Calculations

The 'power tranformed Hartmann distance' are calulated as follows: The simulated Hartmann distribution is added a constant as the Box-Cox transformation can only be applied to positive values. Then a range of values for lambda in the Box-Cox transformation (Box & Cox, 1964) are tried out. The best lambda is the one maximizing the correlation of the quantiles with the standard normal distribution. The lambda value maximizing normality is used to transform Hartmann distances. As the resulting scale of the power transformation depends on lambda, the resulting values are z-transformed to derive a common scaling.

The code for the calculation of the optimal lambda was written by Ioannis Kosmidis.

Details

Hartmann (1992) suggested a transformation of Slater (1977) distances to make them independent from the size of a grid. Hartmann distances are supposed to yield stable cutoff values used to determine 'significance' of inter-element distances. It can be shown that Hartmann distances are still affected by grid parameters like size and the range of the rating scale used (Heckmann, 2012). The function distanceNormalize applies a Box-Cox (1964) transformation to the Hartmann distances in order to remove the skew of the Hartmann distance distribution. The normalized values show to have more stable cutoffs (quantiles) and better properties for comparison across grids of different size and scale range. The function distanceNormalize can also return the quantiles of the sample distribution and only the element distances consideres 'significant' according to the quantiles defined.

References

Box, G. E. P., & Cox, D. R. (1964). An Analysis of Transformations. Journal of the Royal Statistical Society. Series B (Methodological), 26(2), 211-252.

Hartmann, A. (1992). Element comparisons in repertory grid technique: Results and consequences of a Monte Carlo study. International Journal of Personal Construct Psychology, 5(1), 41-56.

Heckmann, M. (2012). Standardizing inter-element distances in grids - A revision of Hartmann's distances, 11th Biennal Conference of the European Personal Construct Association (EPCA), Dublin, Ireland, Paper presentation, July 2012.

Slater, P. (1977). The measurement of intrapersonal space by Grid technique. London: Wiley.

See Also

distanceHartmann and distanceSlater.

Examples

### basics  ###

   distanceNormalized(bell2010)
   n <- distanceNormalized(bell2010)
   n

   # printing options
   print(n)
   print(n, digits=4)
   # 'significant' distances only
   print(n, p=c(.05, .95))

   # access cells of distance matrix
   n[1,2]

   ### advanced  ###

   # histogram of Slater distances and indifference region
   n <- distanceNormalized(bell2010, distributions=TRUE)
   l <- attr(n, "distributions")
   hist(l$bc, breaks=100)