The validity, reliability, & sensitivity of Lectical^™ Assessments*

Validity

When thinking about the validity of our assessments, we focus primarily upon construct validity**. Construct validity is usually thought of as the degree to which assessments measure what they are designed to measure, but since our assessments do much more than provide a score, we think of construct validity more broadly—as the degree to which our assessments accomplish what they are designed to accomplish. When we design and deliver assessments, we consider (1) the extent to which their scores capture the Lectical^™ dimension (the skill level of the performance); (2) how well they target the domain or topic of interest; (3) their relevance, particularly with respect to the relevance of their feedback; and (4) their utility, particularly with respect to the value of their feedback.

Figure 1: Validity considerations for Lectical Assessments and DiscoTests

Reliability

We track two forms of reliability: (1) internal consistency, which we examine with Rasch modeling software, and (2) inter-rater reliability.

Figure 2: Maintaining the reliability of our scores

Research

We have been conducting research on (and with) our assessments for several years. Some of this research has been published in peer reviewed journals. Other research is documented in reports. The following section shows how some of our publications and reports relate to various aspects of reliability and validity. (Click on titles to view pdf documents.)

Predictive validity 1	Evidence
LectaTests are designed to target real-world skills—skills that make us better at what we do at work and in our personal lives. If we are doing a good job, working with our assessments should support behavioral change.	In a preliminary analysis of Clear Impact's ambitious 40-hour, 9 month, leadership training initiative involving four levels of management in a large North American city, we examined the effects of embedding up to 8 LectaTests (including pre-and post LDMAs) on manager's growth and collaborative behavior. Most of the results reported here are restricted to the LDMA data of supervisors who (1) completed pre and post LDMAs and (2) had two or more supervisees who had completed pre and post LDMAs.
	The number of LectaTests taken by supervisors predicts their own Lectical growth 161 Greater Lectical growth of supervisors predicts higher 360 scores from direct reports Regression results   n r p .19 .01 9 .59 .09 Higher 360 scores for supervisors predict higher average lectical growth of direct reports 10 .60 .07 Higher average Lectical growth of direct reports predicts higher 360 scores from peers 10 .46 .18

Predictive validity 2
Upper level managers, on average, have higher-level decision making skills than lower level managers.

Embedding LectaTests

This table shows growth during a current leadership training program in which LectaTests are embedded Embedded assessments inform instruction. For example, they may be used as part of a lesson plan, like a written assignment might be used to help learners solidify what they are learning. as formative assessments. Although class time is limited to 40 hours, participants are demonstrating development that is equivalent to what we have measured in year-long college courses (without embedding).

Detecting change over short periods and in small samples

In the table below, paired samples t-tests show levels of detectable growth in several program evaluations that were conducted with Lectical Assessments. They demonstrate that our measures can detect growth as small as .05 of a level in an average-sized classroom (NA1 2013, LDMA). Moreover, they show that measurable growth can occur with minimal instruction in a well-designed program. For example, the individuals in CI 2012 met only 4 times over the course of 3 months.

*All Lectical assessments meet or exceed the validity and reliability standards for educational and psychological testing set jointly by the APA, AERA, and NCME.

**See: Messick, S. (1980). Test validity and the ethics of assessment. American Psychologist, 35(11), 1012-1027.