The data that I have has strong negative correlation with the response variable.
var1 -0.3049408
var2 -0.2874333
var3 -0.2760590
var4 -0.2708325
var5 -0.2697072
var6 -0.2667927
var7 -0.2425371
var8 -0.2406525
var9 -0.2405309
var10 -0.2263708
But when I added them to the linear models, some of the variables coefficients are positive. My colleague is saying that this doesn't make sense because the correlations are negative and explaining that var 8 has positive coefficients while other variables are -ve doesn't make sense. Is this true? If the model is correct, does it need for the coefficients to be negative if correlations are negative?
