Gravar-mail: Within- and between-group regression for improving the robustness of causal claims in cross-sectional analysis