# PCI1 vs PCI2

PCI2 and associated enhancements offer multiple advantages compared to the initial formulation (PCI1).

• PCI1 required bipolar scales (e.g., -3 to +3) with a neutral point (0) and scale widths were constrained to 3, 5, 7, and 9. PCI2 can be used with scale widths of 2, 3, 4, 5, 6, 7, 8, and 9, and can be applied to bipolar scales (with or without a neutral value) and unipolar scales (e.g., not at all important to extremely important). This expanded flexibility of PCI2allows researchers to use the statistic with virtually all fixed length scales used in survey research.
• PCI2 formulation allows for an indefinite number of distance functions. Current implementation includes two distance formulations (D1 and D2) for bipolar scales and a distance function for unipolar scales (D3). Given that people with negative or positive responses may perceive no disagreement or conflict with a person who is neutral on a topic, D1 does not include neutral responses in the calculation of distance. The second distance function, D2, includes neutral responses when calculating the PCI2. Although D1 is recommended for bipolar scales, the PCI2 allows researchers to experiment with impacts of excluding (D1) or including (D2) “neutral” or “neither” values when calculating the PCI2. For unipolar scales, D3 should be used.
• Because distances need not be linear functions of responses. PCI2 allows for powers of distances. Researchers can examine the impact of linear and non-linear response patterns using the PCI2 distributions generated by the simulation routines. Power 1 (i.e., P1, the unsquared version) is currently recommended for use unless there is a rational for increasing weight as differences between responses become more extreme (i.e., P2, the squared version). Given that any power > 0 can used, options for transforming a distance function are infinite
• The original formulation of the PCI1 required users to calculate a variable’s frequency distribution in a statistical program (e.g., SAS, SPSS) and then type or copy the distribution into Microsoft Excel. The second generation of this statistic allows researchers to produce the statistic directly from SAS or SPSS, or use an Excel spreadsheet.
• The SAS, SPSS and Excel programs also generate a simulation based on a variable’s actual distribution of PCI2; PCI1 did not include such routines. This simulation produces a mean and standard deviation for the estimated PCI2. The standard deviations from the simulated PCI2 is critical in testing for a significant difference from a value (e.g., from 0 or 1) and for testing for significant differences among PCI2 values. The simulation is based on an n of 400, but this default can be changed to any sample size.