Video 46 - Developing Research Hypotheses - ESTIEM LSS Course

Understanding Research Hypotheses and Statistical Testing

Introduction to Research Hypotheses

• The discussion begins with the importance of understanding differences in data, specifically between two subjects (Bob and another individual).

• A research hypothesis is defined as a tentative statement about reality that guides investigation, distinct from a statistical hypothesis which mathematically expresses a test for confidence levels.

Types of Hypotheses

• The speaker introduces "Theory O" or "Theory Opinion," which encompasses personal conjectures or desires that may not have empirical support.

• Emphasis is placed on converting these theories into logical hypotheses to be tested for truth, highlighting the principle of falsification in scientific inquiry.

Process of Falsification

• The process involves evaluating hypotheses by attempting to falsify them; if an alternative can be proven false, it suggests the original hypothesis may hold true.

• Key factors influencing this process include understanding physical aspects, choosing appropriate variables, and ensuring measurement quality.

Characterization and Testing

• The initial step involves breaking down problems into questions leading to theory development; multiple theories can exist for one cause.

• Each theory must define its hypothesis clearly: what is expected to happen versus what is not expected (the alternative hypothesis).

Practical vs. Statistical Significance

• Understanding hypotheses requires examining both practical implications of rejecting a hypothesis and the statistical significance based on observed data adequacy.

• Typical hypotheses are framed around process inputs and outputs, assessing whether outcomes meet decision criteria for success.

Evaluating Process Changes

• Various tests are introduced:

• One-sample T-test compares current outcomes against historical performance.

• Two-sample T-test evaluates changes before and after a process modification.

• One-sample P-test assesses failure rates against budgeted expectations (e.g., warranty rates).

Reliability Assessment