Statistical modelling

We are now ready to fit statistical models to compare tsetse control and non-control periods. We shall apply the model:

y_ijk = µ + s_i + p _j + c_k + (pc)_jk +e_ijk

where y_ijk is the dependent variable, µ = mean, s_i= season (i =1,2), p_j = period (j=1,2), c_k = non-tsetse versus tsetse control (k=1,2), (pc)_jk = interaction between period and effect of tsetse control and e_ijk is the residual or error term.

We shall start with adult male body weight and use the lagged values under the tsetse control column in CS10Data5. When fitting the above model with Stats Regression analysis Generalized Linear Models... we find that the interaction (Period.Lagged) is non-significant. Repeating the analysis without the interaction and clicking the 'Options' button followed by 'Accumulated' we find that the effect of tsetse control is significant (P<0.001)