Unit 3 Quick Info

Two essential features of all statistically designed experiments...

1) Compare several treatments 2) Use chance to assign subjects to treatments

nonsampling error sources

Anything other than "the sample used in the study is not representative of the whole population" Examples are selection bias, population mis-specification error, sampling frame error, processing error, respondent error, non-response error, instrument error, interviewer error, and surrogate error.

Blocking

Blocking is used to separate the experimental units into categories BEFORE an experiment- based on some way that is expected to effect the response to the treatments.

Matched pair example

Each person is considered a pair of sneakers and is given BOTH types of materials, with the type of material randomly assigned to the left or right foot (It would be a completely randomized design if the type of material was randomly assigned to each person).

Note: If you keep a doctor/physician unaware in a double-blind study it...

Eliminates a possible source of bias.

Random assignment in experiments- it's importance

It is a good way to create groups of subjects that are roughly equivalent at the beginning of the experiment.

Random selection of experimental units

Not required, only random ASSIGNMENT of treatments to experimental units is required.

con of observational studies

Observational studies are not enough to confirm causation, an experiment has to be conducted.

Cluster vs. Stratified Sampling

Randomly selected clusters vs. randomly selected members from chosen strata. Homogeneity externally vs. homogeneity internally.

No replication example

The experimental units are ten shrubs, and the treatments are ten pruning methods. Each pruning method was only used on one shrub. Therefore there is no replication of experimental units in each treatment group.

Statistical significance

The results are too large to be attributed to chance alone. The treatments were randomly assigned.

Factors

the explanatory variables in an experiment. Each factor has two or more levels (i.e., different values of the factor). Combinations of factor levels are called treatments.

Experimental units

the smallest COLLECTION of individuals that receive various different treatments.

An experiment compared the adhesion of 2 types of paint, A and B, to 3 types of metal, 1, 2, and 3, used in automobiles. Thirty sheets of metal were used in the experiment. 10 of each type of metal. Half of each metal type will receive paint A and the other half will receive paint B. How many experimental units are in the experiment?

•What I got wrong (you put 3): The number of blocks is 3. The blocks are the type of metal. •Right answer: 30. This is the number of experimental units because it is the total number of sheets to which the treatments (type of paint) are applied.