Type: Package
Title: Magnitude-Based Inferences
Version: 1.3.5
Date: 2019-01-22
Maintainer: Kyle Peterson <petersonkdon@gmail.com>
Description: Allows practitioners and researchers a wholesale approach for deriving magnitude-based inferences from raw data. A major goal of 'mbir' is to programmatically detect appropriate statistical tests to run in lieu of relying on practitioners to determine correct stepwise procedures independently.
Imports: graphics, stats, utils, effsize, psych
URL: http://mbir-project.us/
License: GPL-2
Copyright: Segments of the package are based upon Will G. Hopkins' work. See vignette and COPYRIGHT file for details.
Encoding: UTF-8
LazyData: true
RoxygenNote: 6.1.0
Suggests: knitr, testthat, rmarkdown
VignetteBuilder: knitr
NeedsCompilation: no
Packaged: 2019-01-22 17:40:10 UTC; ath_sportsci
Author: Kyle Peterson [aut, cre], Aaron Caldwell [aut]
Repository: CRAN
Date/Publication: 2019-01-22 17:50:02 UTC

Accuracy in Parameter Estimation: Standardized Mean Difference

Description

Estimates sample size for paired or independent, two-sample study desings via Accuracy in Parameter Estimation. Calculates n so a given study is likely to obtain margin of error no larger than chosen target margin of error.

Usage

aipe_smd(moe, paired = c(TRUE, FALSE), conf.int, assur.lvl, r)

Arguments

moe

target margin of error in standard deviation units

paired

(character) logical indicator specifying if x and y are paired (TRUE) or independent (FALSE)

conf.int

(optional) confidence level of the interval. Defaults to 0.90

assur.lvl

(optional) desired level of assurance (percent experiments whose MOE is less than target MOE). Defaults to 0.99

r

(required if paired = TRUE) population correlation between the two measures

Details

Refer to vignette for further information.

References

Maxwell SE, Kelley K & Rausch JR. (2008). Sample size planning for statistical power and accuracy in parameter estimation. Annual Review of Psychology, 59, 537-563.

Kelley K & Rausch JR. (2006). Sample size planning for the standardized mean difference: Accuracy in parameter estimation via narrow confidence intervals. Psychological Methods, 11, 363–385.

Examples

aipe_smd(moe = 0.55, paired = TRUE, conf.int = .9, assur.lvl = .99, r = 0.75)

Bootstrap Confidence Intervals via Resampling

Description

Provides nonparametric confidence intervals via percentile-based resampling.

Usage

boot_test(x, y, conf.int, resample, med)

Arguments

x, y

numeric vectors of data values

conf.int

(optional) confidence level of the interval. Defaults to 0.90

resample

(optional) number of resamples. Defaults to 10,000

med

(optional) number indicating true difference in medians to test against. Defaults to zero.

Details

Refer to vignette for further information.

Examples

require(graphics)

a <- rnorm(25, 80, 35)
b <- rnorm(25, 100, 50)

boot_test(a, b, 0.95, 10000)

Correlation Coefficient

Description

Provides magnitude-based inferences upon given r value and sample size. Based upon WG Hopkins Microsoft Excel spreadsheet.

Usage

corr(r, n, conf.int = 0.9, swc = 0.1, plot = FALSE)

Arguments

r

correlation coefficient

n

sample size

conf.int

(optional) confidence level of the interval. Defaults to 0.90

swc

(optional) number indicating smallest worthwhile change. Defaults to 0.1

plot

(optional) logical indicator specifying to print associated plot. Defaults to FALSE

Details

Refer to vignette for further information.

References

Hopkins WG. (2007). A spreadsheet for deriving a confidence interval, mechanistic inference and clinical inference from a p value. Sportscience 11, 16-20. sportsci.org/2007/wghinf.htm

Examples

corr(.40, 25, 0.95)

Test of Two Correlations

Description

Provides statistical inference upon the difference between two independent correlations.

Usage

corr_diff(r1, n1, r2, n2, conf.int = 0.9, plot = FALSE)

Arguments

r1

correlation of group 1

n1

sample size of group 1

r2

correlation of group 2

n2

sample size of group 2

conf.int

(optional) confidence level of the interval. Defaults to 0.90

plot

(optional) logical indicator specifying to print associated plot. Defaults to FALSE

Details

Refer to vignette for further information.

References

Zou GY. (2007). Toward using confidence intervals to compare correlations. Psychological Methods, 12, 399-413.

Examples

corr_diff(r1 = 0.20, n1 = 71, r2 = 0.55, n2 = 46)

Correlation Coefficient Test

Description

Provides magnitude-based inferences for the association between given data vectors. Evaluates normality assumption, performs either Pearson or Spearman correlation and subsequently estimates magnitude-based inferences.

Usage

corr_test(x, y, conf.int = 0.9, auto = TRUE, method = "pearson",
  swc = 0.1, plot = FALSE)

Arguments

x, y

numeric vectors of data values

conf.int

(optional) confidence level of the interval. Defaults to 0.90

auto

(character) logical indicator specifying if user wants function to programmatically detect statistical procedures. Defaults to TRUE

method

(character) if auto = F, logical indicator specifying which correlation to execute (pearson, spearman, kendall). Defaults to "pearson".

swc

(optional) number indicating smallest worthwhile change. Defaults to 0.1

plot

(optional) logical indicator specifying to print associated plot. Defaults to FALSE

Details

Refer to vignette for further information.

Value

Associated effect size measure, r, and respective confidence intervals.

Examples

a <- rnorm(25, 80, 35)
b <- rnorm(25, 100, 35)

corr_test(a, b, 0.95)

Effect Size Converter

Description

Converts between equivalent effect size measures: d, r, odds ratio.

Usage

es_convert(x, from = c("d", "or", "r"), to = c("d", "or", "r"))

Arguments

x

numeric value

from

(character) current effect size of x

to

(character) effect size measure to convert to

Details

Refer to vignette for further information.

References

Rosenthal R. (1994). Parametric measures of effect size. In H. Cooper & LV. Hedges (Eds.), The Handbook of Research Synthesis. New York, NY: Sage.

Borenstein M, Hedges LV, Higgins JPT & Rothstein HR. (2009). Introduction to Meta-Analysis. Chichester, West Sussex, UK: Wiley.

Examples

# Odds ratio to Cohen's d
es_convert(1.25, from = "or", to = "d")

Odds Ratio

Description

Provides magnitude-based inferences upon given odds ratio and p-value. Based upon WG Hopkins Microsoft Excel spreadsheet.

Usage

odds(or, p, conf.int = 0.9)

Arguments

or

odds ratio

p

associated p-value

conf.int

(optional) confidence level of the interval. Defaults to 0.90

Details

Refer to vignette for further information.

References

Hopkins WG. (2007). A spreadsheet for deriving a confidence interval, mechanistic inference and clinical inference from a p value. Sportscience 11, 16-20. sportsci.org/2007/wghinf.htm

Examples

odds(1.25, 0.06, 0.95)

Test of Two Proportions

Description

Provides magnitude-based inferences upon given proportions and sample sizes. Based upon WG Hopkins Microsoft Excel spreadsheet.

Usage

prop(p1, n1, p2, n2, conf.int)

Arguments

p1

proportion of group 1

n1

sample size of group 1

p2

proportion of group 2

n2

sample size of group 2

conf.int

(optional) confidence level of the interval. Defaults to 0.90

Details

Refer to vignette for further information.

References

Hopkins WG. (2007). A spreadsheet for deriving a confidence interval, mechanistic inference and clinical inference from a p value. Sportscience 11, 16-20. sportsci.org/2007/wghinf.htm

Examples

prop(p1 = 0.7, n1 = 25, p2 = 0.5, n2 = 20)

Standardized Mean Difference

Description

Provides magnitude-based inferences upon given d, p-value, and degrees of freedom. Based upon WG Hopkins Microsoft Excel spreadsheet.

Usage

smd(es, p, df, conf.int = 0.9, swc = 0.5, plot = FALSE)

Arguments

es

effect size measure (Cohen's d)

p

associated p-value from t-statistic

df

associated degrees of freedom from t-statistic

conf.int

(optional) confidence level of the interval. Defaults to 0.90

swc

(optional) number indicating smallest worthwhile change. Defaults to 0.5

plot

(optional) logical indicator specifying to print associated plot. Defaults to FALSE

Details

Refer to vignette for further information.

References

Hopkins WG. (2007). A spreadsheet for deriving a confidence interval, mechanistic inference and clinical inference from a p value. Sportscience 11, 16-20. sportsci.org/2007/wghinf.htm

Examples

smd(.75, 0.06, 20, 0.95)

Standardized Mean Difference Test

Description

Performs two-sample difference of means analysis to produce magnitude-based inferences. Evaluates both normality and homogeneity, performs either t-test or wilcoxon test, computes effect sizes and estimates magnitude-based inferences. Allows both independent and paired designs.

Usage

smd_test(x, y, paired = c(TRUE, FALSE), auto = TRUE, var = TRUE,
  normal = TRUE, conf.int = 0.9, mu = 0, swc = 0.5, plot = FALSE)

Arguments

x, y

numeric vectors of data values

paired

(character) logical indicator specifying if x and y are paired (TRUE) or independent (FALSE)

auto

(character) logical indicator specifying if user wants function to programmatically detect statistical procedures. Defaults to TRUE

var

(optional) if auto = F, logical indicator specifying if homogeneity of variance assumed. Defaults to TRUE

normal

(optional) if auto = F, logical indicator specifying if normality assumed. Defaults to TRUE

conf.int

(optional) confidence level of the interval. Defaults to 0.90

mu

(optional) number indicating true difference in means to test against. Defaults to zero.

swc

(optional) number indicating smallest worthwhile change. Defaults to 0.5

plot

(optional) logical indicator specifying to print associated plot. Defaults to FALSE

Details

Refer to vignette for further information.

Value

Associated effect size measures (d, r, odds ratio) and respective confidence intervals based upon which statistical test(s) performed.

Examples

a <- rnorm(25, 80, 35)
b <- rnorm(25, 100, 50)

smd_test(a, b, paired = FALSE, conf.int=0.95)

Sample Size Estimation: Correlation Coefficient

Description

Estimates magnitude-based inferences upon planned sample size and r value. Based upon WG Hopkins Microsoft Excel spreadsheet.

Usage

ss_corr(n, r)

Arguments

n

planned sample size

r

planned correlation coefficient

Details

Refer to vignette for further information.

References

Hopkins WG. (2006). Estimating sample size for magnitude-based inferences. Sportscience 10, 63-70. sportsci.org/2006/wghss.htm

Examples

ss_corr(n = 20, r = 0.2)

Sample Size Estimation: Odds Ratio

Description

Estimates magnitude-based inferences upon planned sample size and odds ratio. Based upon WG Hopkins Microsoft Excel spreadsheet.

Usage

ss_odds(exp, con, or)

Arguments

exp

planned sample size of experimental group

con

planned sample size of control group

or

planned odds ratio

Details

Refer to vignette for further information.

References

Hopkins WG. (2006). Estimating sample size for magnitude-based inferences. Sportscience 10, 63-70. sportsci.org/2006/wghss.htm

Examples

ss_odds(exp = 15, con = 18, or = 3.25)

Sample Size Estimation: Standardized Mean Difference

Description

Estimates magnitude-based inferences upon planned sample size and d value. Based upon WG Hopkins Microsoft Excel spreadsheet.

Usage

ss_smd(exp, con, es)

Arguments

exp

planned sample size of experimental group

con

planned sample size of control group

es

planned Cohen's d

Details

Refer to vignette for further information.

References

Hopkins WG. (2006). Estimating sample size for magnitude-based inferences. Sportscience 10, 63-70. sportsci.org/2006/wghss.htm

Examples

ss_smd(exp = 20, con = 15, es = 0.6)

Smallest Worthwhile Change: Individual

Description

Provides longitudinal magnitude-based inferences for an individual's change from previous time point and magnitude of deviation from trend line.

Usage

swc_ind(x, swc, type = c("previous", "trend"), ts, te, main, xlab, ylab)

Arguments

x

numeric vectors of data values

swc

smallest worthwhile change

type

(character) indicator specifying which type of analysis: "previous" or "trend"

ts

(required if type = "trend") target slope

te

(optional) typical error. Defaults to typical error of the estimate

main

(optional) plot title. Defaults to blank

xlab

(optional) x-axis label. Defaults to "Measurement"

ylab

(optional) y-axis label. Defaults to name of x

Details

Refer to vignette for further information.

References

Hopkins WG. (2017). A spreadsheet for monitoring an individual's changes and trend. Sportscience 21, 5-9. sportsci.org/2017/wghtrend.htm

Examples

df<-c(97.5,99.9,100.2,101,101.2,99.8)

swc_ind(x = df, swc = 0.5, te = 1, ts = 0.25, type = "trend")