Frameworks
Receptiviti’s measures can be separated into two categories: proportional measures and normed measures. This is an important distinction for those who are looking to combine and compare multiple measures for the purpose of extracting insights.
Proportional Measures
These measures produce proportion-based scores, meaning the score output represents the portion of the analyzed text that consists of words related to the psychological construct being measured.
The LIWC, LIWC Extension, Emotions (SALLEE), Temporal and Orientation, and Toxicity frameworks all contain proportional measures: for each submitted text sample, Receptiviti analyzes one word at a time. As each word is processed, the dictionary file is searched by category, looking for a category match with the current word. If the target word is matched with a category word, the appropriate word category scale (or scales) for that word is incremented. While SALLEE operates slightly differently than LIWC, LIWC Extension, Temporal and Orientation, and Cognition, these six frameworks count words in a similar fashion. Proportional measures will always provide scores in a range of 0 to 1, except for SALLEE’s sentiment measures, which will always fall between -1.0 to +1.0.
Proportional Measures are not normalized or baselined against a dataset or population. Rather, they produce raw scores.
Toxicity's toxicity_measures are proportional but toxicity_likelihood are likelihoods in the 0 to 1 range.
Normed Measures
The Big 5 Personality, Social Dynamics, Drives, Cognition, Needs and Values, Interpersonal Circumplex, Fast and Slow Thinking Index, and DISC frameworks all contain normed measures. Normed measures are algorithms that have building block components (i.e., ingredients) that contribute to the psychological phenomenon being measured. Our proportional measure frameworks (i.e., LIWC, SALLEE, etc.) are primary sources of ingredients for the algorithms. Some measures are comprised of one component, while other measures are comprised of multiple components.
A minimum of 350 words per analyzed text sample is required to produce valid scores when analyzing text using normed measures.
By default, scores for these measures are normed using Z-scoring. Z-scoring transforms raw scores into standardized scores that show how far a value is from the mean, measured in standard deviations. For example, if someone scores a Z-score of 2 on a psychometric test, it means their score is two standard deviations above the average for that test. This results in a normal distribution. Scores are then projected onto a range from 0 to 100.
A normed score of 80 indicates that the sample is 2.4 standard deviations away from the mean of the norming dataset. Users can choose from Receptiviti's Spoken or Written norming datasets, or create a custom one.
For those who would prefer percentiles, normed measures can alternatively be normed using Rank Norming. Rank Norming adjusts scores by ranking them within a group and then converting those ranks into a new scale, allowing for comparisons across different groups.
A percentile score of 80 indicates that 80% of the samples in the norming dataset — Receptiviti curated or custom — have scores lower than the analyzed language sample.
Please reach out to [email protected] if you would like to adjust your output from default Z-scoring to rank normed percentiles.