Computational thinking (CT) has been advocated as an essential problem solving skill students need to develop. Emphasizing on CT applied in both programming and everyday contexts, we developed a humanoid robotics curriculum and a computerized assessment instrument. We implemented the curriculum with six classes of 125 fifth graders. Quantitative methods were used to compare students' performance from pretest to posttest. Learning analytics techniques were applied to examine students' problem solving processes. The results showed that students' CT performance improved in both programming and everyday reasoning contexts and that the curriculum benefited students with varied initial performance. The study shed light on how to connect and assess CT in everyday reasoning and programming contexts.