Mismatch in sign of partial and semi partial correlations

[aravindhebbali]

Mismatch in the sign of partial and semi partial correlation in ols_correlations().

library(tidyverse)
library(olsrr)
library(ppcor)

# read data
fire <- read_csv("firedata.csv")

lm2 <- lm(damage ~ fighters + severe, data=fire)
ols_regress(lm2)
##                             Model Summary                             
## ---------------------------------------------------------------------
## R                       0.908       RMSE                   20958.007 
## R-Squared               0.824       Coef. Var                 17.899 
## Adj. R-Squared          0.824       MSE                439238058.090 
## Pred R-Squared          0.823       MAE                    16952.542 
## ---------------------------------------------------------------------
##  RMSE: Root Mean Square Error 
##  MSE: Mean Square Error 
##  MAE: Mean Absolute Error 
## 
##                                       ANOVA                                       
## ---------------------------------------------------------------------------------
##                         Sum of                                                   
##                        Squares         DF      Mean Square       F          Sig. 
## ---------------------------------------------------------------------------------
## Regression        2.050264e+12          2     1.025132e+12    2333.887    0.0000 
## Residual      437920343915.371        997    439238058.090                       
## Total             2.488184e+12        999                                        
## ---------------------------------------------------------------------------------
## 
##                                          Parameter Estimates                                          
## -----------------------------------------------------------------------------------------------------
##       model          Beta    Std. Error    Std. Beta       t        Sig          lower         upper 
## -----------------------------------------------------------------------------------------------------
## (Intercept)     55253.771      1669.994                  33.086    0.000     51976.665     58530.877 
##    fighters    -12711.639       552.040       -0.922    -23.027    0.000    -13794.932    -11628.345 
##      severe     29951.971       695.578        1.725     43.061    0.000     28587.007     31316.935 
## -----------------------------------------------------------------------------------------------------

pcor.test(fire$damage,fire$fighters,fire[,c("severe")])
##     estimate     p.value statistic    n gp  Method
## 1 -0.5892224 2.01807e-94 -23.02667 1000  1 pearson

spcor.test(fire$damage,fire$fighters,fire[,c("severe")])
##     estimate      p.value statistic    n gp  Method
## 1 -0.3059425 4.330228e-23 -10.14676 1000  1 pearson

pcor.test(fire$damage,fire$severe,fire[,c("fighters")])
##    estimate      p.value statistic    n gp  Method
## 1 0.8064269 1.03021e-229  43.06058 1000  1 pearson

spcor.test(fire$damage,fire$severe,fire[,c("fighters")])
##    estimate      p.value statistic    n gp  Method
## 1 0.5721218 6.390856e-88  22.02586 1000  1 pearson

ols_correlations(lm2)
##                Correlations                
## ------------------------------------------
## Variable    Zero Order    Partial    Part  
## ------------------------------------------
## fighters         0.705      0.589    0.306 
## severe           0.855      0.806    0.572 
## ------------------------------------------