NSES Content Standard Unifying Concepts and Processes:  
Evidence, models and explanation 
Grades 9-12, page 117 

Evidence consists of observations and data on which to base scientific explanations. Using evidence to understand interactions allows individuals to predict changes in natural and designed systems.
 

 
Benchmark 9D The Mathematical World: Uncertainty
Grades 3-5, page 227
Some predictions can be based on what is known about the past, assuming that conditions are pretty much the same now.

Benchmark 9D The Mathematical World: Uncertainty
Grades 6-8, page 229
How probability is estimated depends on what is known about the situation. Estimates can be based on data from similar conditions in the past or on the assumption that all the possibilities are known.

Benchmark 9D The Mathematical World: Uncertainty
Grades 9-12, page 230
Even when there are plentiful data, it may not be obvious what mathematical model to use to make predictions from them or there may be insufficient computing power to use some models.

Benchmark 11B Common Themes: Models
Grades 9-12, page 270
The basic idea of mathematical modeling is to find a mathematical relationship that behaves in the same ways as the objects or processes under investigation. A mathematical model may give insight about how something really works or may fit observations very well without any intuitive meaning.

Science for All Americans The Nature of Science
Chapter 1, page 6
Scientists strive to make sense of observations of phenomena by constructing explanations for them that use, or are consistent with, currently accepted scientific principles. Such explanations theories may be either sweeping or restricted, but they must be logically sound and incorporate a significant body of scientifically valid observations.

Science for All Americans The Nature of Science
Chapter 1, page 6
The essence of science is validation by observation. But it is not enough for scientific theories to fit only the observations that are already known. Theories should also fit additional observations that were not used in formulating the theories in the first place; that is, theories should have predictive power.