**Heatwave effect Estimation via Adjustment for Temperature****(HEAT)** is an *R* package that allows researchers to use a novel method to estimate the effect of heat waves. This novel method estimates the effect of heat waves based on a characterization of the effect of heat waves, including well-defining a counterfactual risk, adjustment for some types of temperature as a confounder, and a careful specification of lagged variables for an indicator of a heatwave and temperatures. Please refer to my paper in the hyperlink above.

*Conditional Generalized Propensity Score-based Spatial Matching* * (CGPSsm) *is a novel generalized propensity score (GPS) matching method that enables epidemiologists to estimate the average treatment effect on the treated (ATT) for continuous exposure conditional on its binary exposure status while adjusting for unmeasured spatial confounding.

*,*

**R package***has been developed (version 1).*

**CGPSspatialmatch**In epidemiological studies,

*exposure of interest*may have multiple dimensions such as

*proximity*and

*other quantity-wise characteristics*.

*Proximity*can be coded as a binary variable indicating whether exposure is located within a specified radius (i.e., “buffer”) from the location of observational units (e.g., subjects). A

*quantity-wise characteristic*is coded as a continuous variable that depends on the binary variable. Examples include:

-the number of available ambulance/fire trucks/ICU beds within 10-km radius from residence of patients

-tree covers within 50-m radius from residence of mothers

-the number of oil and gas wells within 10-km radius from residence of subjects

-the level of emissions from facilities within 5-km residence of subjects

These exposure metrics are

*continuous*but

*defined as 0*if there does not exist exposure within a “buffer”.

It is common that

*exposure of interest*and

*a health outcome of interest*exhibit

*spatial patterns*such that

*unmeasured spatial confounding may be concerning*. One may want to use traditional GPS-based methods and use spatial regression models to estimate GPS by treating such types of exposure as just a continuous variable. However, these approaches may be problematic because such types of exposure may be skewed/bimodal-distributed. GPS estimation requires appropriate consideration of exposure distribution.

In this landscape,

*and*

**CGPSsm***have been developed to overcome these unique challenges. CGPSsm maintains the salient benefits of PS matching and spatial analysis:*

**CGPSspatialmatch***straightforward assessments of covariate balance*and

*adjustment for unmeasured spatial confounding*.

Home | About | Publications | R Codes | R Packages