# AP Statistics Curriculum 2007 IntroDesign

### From Socr

(→Approach: added Placebo section) |
|||

(23 intermediate revisions not shown) | |||

Line 1: | Line 1: | ||

- | + | [[AP_Statistics_Curriculum_2007 | General Advance-Placement (AP) Statistics Curriculum]] - Design and Experiments | |

- | + | ==Design of Experiments== | |

''Design of experiments'' refers of the blueprint for planning a study or experiment, performing the data collection protocol and controlling the study parameters for accuracy and consistency. Design of experiments only makes sense in studies where variation, chance and uncertainly are present and unavoidable. Data, or information, is typically collected in regard to a specific process or phenomenon being studied to investigate the effects of some controlled variables (independent variables or predictors) on other observed measurements (responses or dependent variables). Both types of variables are associated with specific observational units (living beings, components, objects, materials, etc.) | ''Design of experiments'' refers of the blueprint for planning a study or experiment, performing the data collection protocol and controlling the study parameters for accuracy and consistency. Design of experiments only makes sense in studies where variation, chance and uncertainly are present and unavoidable. Data, or information, is typically collected in regard to a specific process or phenomenon being studied to investigate the effects of some controlled variables (independent variables or predictors) on other observed measurements (responses or dependent variables). Both types of variables are associated with specific observational units (living beings, components, objects, materials, etc.) | ||

- | + | ==Approach== | |

- | The following are the most common components used in Experimental Design | + | The following are the most common components used in Experimental Design: |

- | + | ===Comparison=== | |

+ | To make inference about effects, associations or predictions, one typically has to compare different groups subjected to distinct conditions. This allows contrasting observed responses and underlying group differences that may ultimately lead to inference on relationships and influences between controlled and observed variables. | ||

- | + | ===Randomization=== | |

+ | The second fundamental design principle is randomization. It requires that we make allocation of (controlled variables) treatments to units using some [[SOCR_EduMaterials_Activities_RNG | random mechanism]]. This simply guarantees that the effects resulted are the units, but not incorporated in the model, equidistributed amongst all groups. Therefore it is unlikely to significantly affect our group comparisons at the end of the statistical inference or analysis (as these effects, if present, will be similar within each group). | ||

- | + | ===Experimental vs. Observational Studies=== | |

+ | There are many situations, where randomized experiments, albeit desirable, are impractical. Therefore, we cannot deduce causality or effects of various treatments on the response measurement. ''Observational studies'' are retrospective or prospective studies where the investigator does not have control over randomization of treatments to subject, or units. In these cases, the subjects or units fall naturally within a treatment group. Examples of such observational studies include [http://www.cancer.org/docroot/PED/ped_10_1.asp studies of smoking effects on cancer] and [[AP_Statistics_Curriculum_2007_IntroDesign#References | use of parachutes to prevent death and major trauma related to gravitational pull]]. | ||

- | + | ===Replication=== | |

+ | All measurements, observations or data collected are subject to variation, as [[SOCR_EduMaterials_Activities_RNG | there are no completely deterministic processes]]. As we try to make inference about the process that generated the observed data (not the sample data itself, even though our statistical analysis is data-driven and based on the observed measurements), the more data we collect (unbiased) the stronger our inference is likely to be. Therefore, repeated measurements intuitively would allow it to tame the variability associated with the phenomenon we study. | ||

- | + | ===Blocking=== | |

+ | Blocking is related to randomization. The difference is that we use blocking when we know ''a priori'' of certain effects of the observational units on the response measurements (e.g., when studying the effects of hormonal treatments on humans, gender plays a significant role). We arrange units into groups (blocks) that are similar to one another when we design an experiment in which certain unit characteristics are known to affect the response measurements. Blocking reduces known and irrelevant sources of variation between units and allows greater precision in the estimation of the source of variation in the study. | ||

- | === | + | ===Orthogonality=== |

- | + | Orthogonality allows division of complex relations and variation into separate (independent/orthogonal) contrasts, or factors, that can be studied efficiently and autonomously. Often, these contrasts may be represented by vectors, where sets of orthogonal contrasts are uncorrelated and may be independently distributed. Independence implies that each orthogonal contrast provides complementary information to other contrasts (i.e., other treatments). The goal is to completely decompose the variance or the relations of the observed measurements into independent components (e.g., like [http://en.wikipedia.org/wiki/Taylor_expansion Taylor expansion] allowing polynomial decomposition of smooth functions, where the polynomial base functions are easy to differentiate, integrate, etc.) This will, of course, allow easier interpretation of the statistical analysis and the findings of the study. | |

- | + | ===Placebo=== | |

+ | A [http://en.wikipedia.org/wiki/Placebo placebo] is a sham or simulated medical intervention that has no direct health impact but may result in actual improvement of a medical condition or disorder. If such sham effect is observed it is called a ''placebo effect''. | ||

- | + | Many health studies use placebos control and measure real effects of newly introduced remedies to assess their efficacy. Common placebos are inert tablets (sugar pills), sham surgery, and other procedures based on false information. An example of a placebo procedure is giving a patient a pill identical to the actual treatment pill, but without the treatment ingredients and measuring the improvement of the patient's condition. Typically all patients are informed that some will be treated using the drug and some will receive the insert pill, however the patients are blinded as to whether they actually received the drug or the placebo. Such an intervention may cause the patient to believe the treatment will change their condition, which may produce a subjective perception of a therapeutic effect. | |

- | + | ||

- | = | + | A [http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0015591 recent Irritable Bowel Syndrome (IBS) study] showed that in some conditions placebo effects may actually be stronger than actual treatment effects. See [http://m.npr.org/news/Health/132281484?singlePage=true this NPR report on this IBS study]. |

- | + | ||

+ | ==Model Validation== | ||

+ | All of the components in the [[AP_Statistics_Curriculum_2007_IntroDesign#Approach | approach/methods section]] need to be validated but the major one is the independence assumption. | ||

+ | ==Computational Resources: Internet-based SOCR Tools== | ||

+ | * [http://www.socr.ucla.edu/htmls/SOCR_ChoiceOfStatisticalTest.html How to choose a statistical test based on a study design?] | ||

+ | |||

+ | ==Examples & hands-on activities== | ||

+ | ===A [[AP_Statistics_Curriculum_2007_IntroDesign#References | study of aortic valve-sparing repair<sup>*</sup>]]=== | ||

+ | * This study sought to establish whether there was a difference in outcome after aortic valve repair with autologous pericardial leaflet extension in acquired versus congenital valvular disease. In this study, 128 patients underwent reparative aortic valve surgery at UCLA from 1997 through 2005 for acquired or congenital aortic valve disease. The acquired group (43/128) (34%) had a mean age of 56.4 <math>\pm</math> 20.3 years (range, 7.8—84.6 years) and the congenital group (85/128) (66%) had a mean age of 16.9 <math>\pm</math>19.2 years (range, 0.3—82 years). The endpoints of the study were mortality and reoperation rates. | ||

+ | * In this case the ''units'' are heart disease patients. These were split into two groups (acquired or congenital) and blocked by gender (male/female). The treatment allied on the two groups was ''aortic valve repair with autologous pericardial leaflet extension''. | ||

+ | |||

+ | ==[[EBook_Problems_EDA_IntroDesign|Problems]]== | ||

<hr> | <hr> | ||

- | + | ||

- | * | + | ==References== |

+ | * [http://www.loni.ucla.edu/~dinov/publications.html David De La Zerda, Oved Cohen, Michael C. Fishbein, Jonah Odim, Carlos A Calderon, Diana Hekmat, Ivo Dinov and Hillel Laks]. [http://www.sciencedirect.com/science?_ob=MImg&_imagekey=B6T35-4MNRN1T-2-7&_cdi=4937&_user=4423&_orig=search&_coverDate=12%2F27%2F2006&_sk=999999999&view=c&wchp=dGLbVlb-zSkWA&md5=e0692b12bb461286315fd4bdf2db5c4f&ie=/sdarticle.pdf Aortic valve-sparing repair with autologous pericardial leaflet extension has a greater early re-operation rate in congenital versus acquired valve disease.] European Journal of Cardio-Thoracic Surgery, February 2007; 31: 256-260 . PMID: 17196393, doi:10.1016/j.ejcts.2006.11.027. | ||

+ | |||

+ | * Gordon Smith, and Jill Pell. (2008) Parachute use to prevent death and major trauma related to gravitational challenge: systematic review of randomised controlled trials. [http://www.bmj.com/cgi/content/full/327/7429/1459?ijkey=425457f110f8db584617b87a1eace92eaa39ff02 BMJ 2003;327:1459-1461 (20 December), doi:10.1136/bmj.327.7429.1459 ] | ||

<hr> | <hr> |

## Current revision as of 19:14, 28 December 2010

General Advance-Placement (AP) Statistics Curriculum - Design and Experiments

## Contents |

## Design of Experiments

*Design of experiments* refers of the blueprint for planning a study or experiment, performing the data collection protocol and controlling the study parameters for accuracy and consistency. Design of experiments only makes sense in studies where variation, chance and uncertainly are present and unavoidable. Data, or information, is typically collected in regard to a specific process or phenomenon being studied to investigate the effects of some controlled variables (independent variables or predictors) on other observed measurements (responses or dependent variables). Both types of variables are associated with specific observational units (living beings, components, objects, materials, etc.)

## Approach

The following are the most common components used in Experimental Design:

### Comparison

To make inference about effects, associations or predictions, one typically has to compare different groups subjected to distinct conditions. This allows contrasting observed responses and underlying group differences that may ultimately lead to inference on relationships and influences between controlled and observed variables.

### Randomization

The second fundamental design principle is randomization. It requires that we make allocation of (controlled variables) treatments to units using some random mechanism. This simply guarantees that the effects resulted are the units, but not incorporated in the model, equidistributed amongst all groups. Therefore it is unlikely to significantly affect our group comparisons at the end of the statistical inference or analysis (as these effects, if present, will be similar within each group).

### Experimental vs. Observational Studies

There are many situations, where randomized experiments, albeit desirable, are impractical. Therefore, we cannot deduce causality or effects of various treatments on the response measurement. *Observational studies* are retrospective or prospective studies where the investigator does not have control over randomization of treatments to subject, or units. In these cases, the subjects or units fall naturally within a treatment group. Examples of such observational studies include studies of smoking effects on cancer and use of parachutes to prevent death and major trauma related to gravitational pull.

### Replication

All measurements, observations or data collected are subject to variation, as there are no completely deterministic processes. As we try to make inference about the process that generated the observed data (not the sample data itself, even though our statistical analysis is data-driven and based on the observed measurements), the more data we collect (unbiased) the stronger our inference is likely to be. Therefore, repeated measurements intuitively would allow it to tame the variability associated with the phenomenon we study.

### Blocking

Blocking is related to randomization. The difference is that we use blocking when we know *a priori* of certain effects of the observational units on the response measurements (e.g., when studying the effects of hormonal treatments on humans, gender plays a significant role). We arrange units into groups (blocks) that are similar to one another when we design an experiment in which certain unit characteristics are known to affect the response measurements. Blocking reduces known and irrelevant sources of variation between units and allows greater precision in the estimation of the source of variation in the study.

### Orthogonality

Orthogonality allows division of complex relations and variation into separate (independent/orthogonal) contrasts, or factors, that can be studied efficiently and autonomously. Often, these contrasts may be represented by vectors, where sets of orthogonal contrasts are uncorrelated and may be independently distributed. Independence implies that each orthogonal contrast provides complementary information to other contrasts (i.e., other treatments). The goal is to completely decompose the variance or the relations of the observed measurements into independent components (e.g., like Taylor expansion allowing polynomial decomposition of smooth functions, where the polynomial base functions are easy to differentiate, integrate, etc.) This will, of course, allow easier interpretation of the statistical analysis and the findings of the study.

### Placebo

A placebo is a sham or simulated medical intervention that has no direct health impact but may result in actual improvement of a medical condition or disorder. If such sham effect is observed it is called a *placebo effect*.

Many health studies use placebos control and measure real effects of newly introduced remedies to assess their efficacy. Common placebos are inert tablets (sugar pills), sham surgery, and other procedures based on false information. An example of a placebo procedure is giving a patient a pill identical to the actual treatment pill, but without the treatment ingredients and measuring the improvement of the patient's condition. Typically all patients are informed that some will be treated using the drug and some will receive the insert pill, however the patients are blinded as to whether they actually received the drug or the placebo. Such an intervention may cause the patient to believe the treatment will change their condition, which may produce a subjective perception of a therapeutic effect.

A recent Irritable Bowel Syndrome (IBS) study showed that in some conditions placebo effects may actually be stronger than actual treatment effects. See this NPR report on this IBS study.

## Model Validation

All of the components in the approach/methods section need to be validated but the major one is the independence assumption.

## Computational Resources: Internet-based SOCR Tools

## Examples & hands-on activities

### A study of aortic valve-sparing repair^{*}

- This study sought to establish whether there was a difference in outcome after aortic valve repair with autologous pericardial leaflet extension in acquired versus congenital valvular disease. In this study, 128 patients underwent reparative aortic valve surgery at UCLA from 1997 through 2005 for acquired or congenital aortic valve disease. The acquired group (43/128) (34%) had a mean age of 56.4 20.3 years (range, 7.8—84.6 years) and the congenital group (85/128) (66%) had a mean age of 16.9 19.2 years (range, 0.3—82 years). The endpoints of the study were mortality and reoperation rates.
- In this case the
*units*are heart disease patients. These were split into two groups (acquired or congenital) and blocked by gender (male/female). The treatment allied on the two groups was*aortic valve repair with autologous pericardial leaflet extension*.

## Problems

## References

- David De La Zerda, Oved Cohen, Michael C. Fishbein, Jonah Odim, Carlos A Calderon, Diana Hekmat, Ivo Dinov and Hillel Laks. Aortic valve-sparing repair with autologous pericardial leaflet extension has a greater early re-operation rate in congenital versus acquired valve disease. European Journal of Cardio-Thoracic Surgery, February 2007; 31: 256-260 . PMID: 17196393, doi:10.1016/j.ejcts.2006.11.027.

- Gordon Smith, and Jill Pell. (2008) Parachute use to prevent death and major trauma related to gravitational challenge: systematic review of randomised controlled trials. BMJ 2003;327:1459-1461 (20 December), doi:10.1136/bmj.327.7429.1459

- SOCR Home page: http://www.socr.ucla.edu

Translate this page: