interactionTest (version 1.0)

findMultiLims: Determine Critical t-Statistic For Marginal Effects Plot

Description

This function determines the appropriate critical t-statistic that limits the false null rejection rate of a marginal effects plot to a specified value, err, using bootstrapped samples of calculated marginal effects values.

Usage

findMultiLims(mat1, mat2 = NULL, mat3 = NULL, mat4 = NULL, p1 = 99, p2 = NULL, p3 = NULL, p4 = NULL, type = "any", err = 0.05)

Arguments

mat1

Matrix of bootstrapped samples of marginal effects.

mat2

Matrix of bootstrapped samples of marginal effects.

mat3

Matrix of bootstrapped samples of marginal effects.

mat4

Matrix of bootstrapped samples of marginal effects.

p1

The type of hypothesis test for the marginal effects in mat1. 1 = one-sided test for ME > 0, -1 = one-sided test for ME < 0, 0 = null of ME = 0 cannot be rejected, any other value = two-sided test for ME != 0

p2

The type of hypothesis test for the marginal effects in mat2. 1 = one-sided test for ME > 0, -1 = one-sided test for ME < 0, 0 = null of ME = 0 cannot be rejected, any other value = two-sided test for ME != 0

p3

The type of hypothesis test for the marginal effects in mat3. 1 = one-sided test for ME > 0, -1 = one-sided test for ME < 0, 0 = null of ME = 0 cannot be rejected, any other value = two-sided test for ME != 0

p4

The type of hypothesis test for the marginal effects in mat4. 1 = one-sided test for ME > 0, -1 = one-sided test for ME < 0, 0 = null of ME = 0 cannot be rejected, any other value = two-sided test for ME != 0

type

The condition that corresponds to successful rejection of the null. "any" indicates any of the matK marginal effects are statistically significant in the direction given by pK, "all" = ALL of the matK marginal effects are statistically significant and in the direction given by pK

err

Rejection rate.

Value

The limits of the t-statistic for the given rejection rate

References

Clark, William R., and Matt Golder. 2006. "Rehabilitating Duverger's Theory." Comparative Political Studies 39(6): 679-708.

Esarey, Justin, and Jane Lawrence Sumner. 2015. "Marginal Effects in Interaction Models: Determining and Controlling the False Positive Rate." URL: http://jee3.web.rice.edu/interaction-overconfidence.pdf.

Examples

## Not run: 
#   data(legfig)                # Clark and Golder 2006 replication data
#   set.seed(1231124)
# 
#   # limit to established democracies from the 1990s
#   dat<-subset(legfig, subset=(nineties==1 & old==1))
# 
#   # create bootstrap samples of marginal effects of eneg and logmag on enep1
#   # uses the bootFun utility included in this package
#   library(boot)
#   boot.t.dist <- boot(data = dat, statistic = bootFun, R = 1000,
#             form=enep1 ~ eneg * logmag + uppertier_eneg + uppertier + proximity1 +
#             proximity1_enpres + enpres, fam="gaussian", x.name="eneg",
#             z.name="logmag")$t
#   boot.t.x.dist<-boot.t.dist[,1:10]
# 
# 
#   # calculate critical t-statistic that sets familywise error rate to 10%
#   # for statistical significance of marginal effect of of eneg at any value of logmag
#   findMultiLims(boot.t.x.dist, type="any", err=0.1)$minimum # answer: 2.593086
# 
#   # calculate critical t-statistic that sets FWER to 10% for ME of eneg = 0
#   # when logmag is small and ME of eneg > 0 when logmag is large
#   boot.t.x.dist.lo<-boot.t.dist[,1:5]
#   boot.t.x.dist.hi<-boot.t.dist[,6:10]
#   findMultiLims(boot.t.x.dist.lo, boot.t.x.dist.hi, type="all", p1=0,
#                   p2=1, err=0.1)$minimum     # answer: 1.008688
# ## End(Not run)