SCI 201 Chapter 2 - Planning and Preparation (Questions)

State the flow of a basic research proposal, including 2 main sections and their subsections.

Title and Title Page

Abstract and Keywords

Introduction:

• Statement of the research problem and objectives of the study

• Literature review

• Study area

• Specific objectives and hypotheses

Materials and Methods:

• Research design/sampling strategy

• Procedures/techniques

• Statistical analysis

• References

• Appendices

Explain how using boolean keywords (or, and) can lead to better searches.

OR: gives broader search results. Searching “jelly” OR “jam” gives results with jelly, jam, or both

AND: gives narrower or more specific search results. Searching “peanut butter” and “jelly” gives results with BOTH peanut butter AND jelly

What are the components of research design that must be included in the proposal and final design?

Study units, variables, spatial aspects, temporal aspects, techniques and methods of data collection, forward planning for data analysis, repeatability, sampling strategies and randomness, overall type of research design (observational vs experimental)

Compare the following variables by numerical value, scale, and measurement types: categorical, ordinal, quantitative.

Categorical: nonnumerically valued, nominal scale, frequencies

Ordinal: nonnumerically valued, ordinal scale, frequencies

Quantitative: numerically valued, ordinal scale, numbers/values

Categorize temperature measured in Kelvin and measured in Celsius as ratio or interval scale. Explain.

Kelvin: ratio scale. 0K is the absolute 0 point of temperature, thus any value of Kelvin is a real measurement of increase in temperature. Ratios can be formed between measurements of temperature in Kelvin.

Celsius: interval scale. 0C is arbitary and does not mean an actual zero point of temperatures, thus any ratios established are invalid

Explain how sampling with and without replacement affects the independence requirement.

With replacement: ensures that each individual has an equal chance of being selected every time

Without replacement: violates the independence requirement because the chance of subsequent individuals being chosen will change (increase as sample size decreases)

Explain how increasing sample size increases power of the test.

Reduces standard error of the estimate (s/sqrt(n)), which increases the value of the test statistic and thus decreases the p-value. Thus, the power of the test increases

Explain how the degree of variability in a sample affects how the subsequent sample size needs to be changed.

Low variability within groups: small sample size is adequate because natural variation is small

High variability within groups: large sample size needed because natural variation is large

Explain how the magnitude of difference in a sample affects how the subsequent sample size needs to be changed.

Low magnitude of difference between groups: would be different to notice an effect between the control/baseline group and the experimental/manipulated group, thus a large sample size is needed to reduce noise

High magnitude of difference between groups: would be easily able to notice an effect difference, smaller sample size is necessary

Larger sample size (increases/decreases) sampling variability. Explain.

Decreases, more of the sample is included/recognized in the chosen sample, so each chosen sample is more similar

Explain the difference between convenience sampling and voluntary response bias, and where their respective source of bias comes from.

Convenience sampling is biased because it does not provide the researcher with a randomized sample that is representative of the population, often favouring certain demographics .

Voluntary response bias is due to the responses not providing a random sample from the target population, and the data is only taken from those who decided to give their response, which typically skews the data to those who favour an outcome strongly.

What are the two most common causes of undercoverage bias?

Incomplete sampling frame, conducting and SRS instead of stratified/population allocation

Describe the key points of difference between experimental and observational studies, including imposing of treatments, manipulation and control of groups, and control of extraneous variables.

Observational: no imposing of treatments between groups, no manipulation or control of groups, no controlling extraneous variables

Experimental: different levels of treatment imposed between groups, at least one control and one experimental group with manipulated treatment, attempt to control extraneous variables by making them constant across all groups

Explain what inferences can be made with the types of studies. What precursor must be met in order for these inferences to be made?

Population: random sampling of study units from the target population required

• Observational studies and experimental studies, if there is random sampling

Causal: random allocation between groups, manipulation of variables, and control of extraneous factors

• Experimental studies, if all conditions are met

• Never observational studies because there is no manipulation, extraneous control, or random allocation

In tests with one or more factors that are all categorical variables, when testing for any associations between observed frequencies and expected frequencies based on a theoretical ratio, which tests should be used in analysis?

Chi squared test of association

In tests with one sample proportion of the population, when testing for any differences between observed proportions and a hypothesized proportion, which tests should be used in analysis?

One proportion z-test (greater than/less than hypothesized)

In tests with two sample proportions of two populations, when testing for any differences between the proportions of the two populations, which tests should be used in analysis?

Two proportion z-test (greater/less than the other sample proportion)

In tests with one explanatory variable with two factors and one quantitative response variable, when testing for any differences in the response variable between groups, which tests should be used in analysis?

Two mean t-test for independent samples (parametric) or the Mann Whitney U-test (nonparametric)

In tests with one explanatory variable with more than two factors and one quantitative response variable, when testing for any differences in the response variable between groups, which tests should be used in analysis?

Single factor ANOVA (parametric) or Kruskal-Wallis test (nonparametric)

In tests with one explanatory variable and one block variable (each more than 1 factor) and one quantitative response variable, when testing for any differences in the response variable between groups, which tests should be used in analysis?

Randomized block ANOVA (parametric) or Friedman’s test (nonparametric)

In tests with two categorical explanatory variables (each more than 1 factor) and one quantitative response variable, when testing for any differences in the response variable between groups and variables, which tests should be used in analysis?

Two factor ANOVA with interaction

In tests with more than two categorical explanatory variables (each more than 1 factor) and one quantitative response variable, when testing for any differences in the response variable between groups and variables, which tests should be used in analysis?

Multi-way factorial ANOVA

In tests with one quantitative explanatory variable and one quantitative response variable, when testing for the effect of the explanatory on the response variable, which tests should be used in analysis?

Simple regression analysis (ANOVA f-test or regression t-test or correlation r-test), nonparametric spearman correlation test

In tests with more than one quantitative explanatory variable and one quantitative response variable, when testing for the effect of several explanatory variables on the response variable, which tests should be used in analysis?

Multiple linear regression (overall regression ANOVA F-test and t-tests for individual slopes and correlation r-tests)

What is the difference between voluntary sampling bias and convenience sampling bias?

The main difference between the two is that in convenience sampling, the researcher is the one actively choosing the participants, while in voluntary sampling the researcher does not have control over who volunteers and is thus chosen. Even though the end product for both methods is a biased sample, voluntary sampling may have originally targeted an unbiased/random sample (that is then no longer random because of those who did not volunteer), while convenience sampling does not take into account randomization in the first place.