Variable strong in Pearson coefficient calculation but weak in linear regression

[Boruchf]

Pearson coefficient calculations:

# Jews v. Cases/Million r = 0.6199 (p = .000154)
# Christians v. Cases/Million r = .5876 (p = .000416)
# Int Adoptions v. Cases/Million r = 0.5371 (p = .000675)
Tourist Revenue v. Cases/Million r = .483 (p= .005)
# Tests/million v. Cases/Million r = 0.3925 (p = .026)
Wealth v. Cases/Million r = 0.3592 (p = .044)

Not significant.
Imports v cases/million
GDP/person v cases per million
Weighted population density v cases/million
Democracy index v cases/million
# Muslims v cases/million
Individuality index v cases/million
Social tightness v cases/million.


Regression equations:

1) Cases/million = 481.8278 + 0.4358 x christians + 1.2311 x jews + 14.168 x tests/million; 481.8278 (p = .5896 ) .4358 (p = .0002 ) 1.2311 (p = .0052 ) 14.1658 ( p = .0004 ) ; overall r6688 .=2 ; p = <.0001
2) cases/million = 811.53656 + .47958 x #christians + .47737 x int. adoptions + 15.16184 x tests/million - .80862 x Net Wealth ; 811.53656 (p = .42816), .47958 (p = .00083), .47737 (p = .04058), 15.16184 (p = .00493), .80862 (p = .10171) ; overall: R- squared .62492, p = .00002.
3) cases/million = 2552.65964 + 1.28701 x #Jews + .49275 x int. adoptions + 8.52382 x tests/million - .56025 x Net Wealth ; 2552.65964 (p = .02), 1.28701 (p = .04), .49275 (p = .07392), 8.52382 (p = .136), -.56025 (p = .30684) ; overall: R- squared .51199, p = .00049
4) cases/million = 1042.7738 + 1.2967 x # Jews + .4841 x # Christians + .3059 x # int. adoptions -10.3612 x tourist revenue + 13.2326 x tests/million - .7981 x wealth ; 1042.7738 (p = .3588), 1.2967 (p = .0135), .4841 (p = .0003), .3059 (p = .3949), -10.3612 (p = .8181), 13.2326 ( p = .0077), .7981 (p = .1083); overall: R-squared = .7137, p = <.0001.

My study suggests a psychological connection between # int. adoptions and COVID19 cases/million; however #int. adoptions doesn't perform well in linear regression equations. What can I conclude?

 

[hlsmith]

It is a little hard to discern what you have going on above (text kind of congested), but did you run a bunch of bivariate correlations then multiple regressions? If so, they are not the same thing, since the latter is 'adjusting' for the other variables in the model at the same time.