Interaction between quantitative predictors

Transcription

1 Interaction between quantitative predictors In a first-order model like the ones we have discussed, the association between E(y) and a predictor x j does not depend on the value of the other predictors in the model. See Fig. 4.1: relation between E(y) and x 1 is the same regardless of the value of x 2 : all the prediction lines are parallel. If, however, the association between response and one of the predictors depends on the value of other predictors, then a first-order model is no longer appropriate. We say that there is an interaction among predictors. Stat Fall

2 Interaction (cont d) Example: a company wishes to estimate the association between sales of a beauty product (y) and two potential predictors of sales in each of n markets: $ spent on daytime TV ads in ith market (x 1 ) and average number of years of education of females in ith market. Intuitively, this is what we would expect: Advertisement expenses will tend to increase sales (up to a point). In cities where women are highly educated (on the average), less of them will be watching TV during the day. The effect of $ in ads on sales may then also depend on education of potential consumers. Stat Fall

3 Interaction (cont d) A figure to represent the association between ads and sales for different levels of education will be drawn in class. How do we include an interaction term in the model? With k = 2 predictors: y i = β 0 + β 1 x 1i + β 2 x 2i + β 3 x 1i x 2i + ɛ i, where the assumptions about the model are the same as before. An interaction between two predictors is a second-order term in the model. Stat Fall

4 Interaction (cont d) In sales example, we would expect that β 3 < 0: as education increases (and more women are out working), the strength of the association between daytime TV ads on sales decreases. In other words, daytime ads are expected to be more effective in markets where more women are at home watching TV during the day than in markets where most women are not watching TV. In general, with k predictors, we can include pairwise interactions between any two, as appropriate. Higher order interactions (e.g. x j x l x t denoting the three-way interaction between the jth, lth and tth predictors) can also be included in the model, but are much harder to interpret from a subject matter point of view. Stat Fall

5 Interaction (cont d) When predictors interact, the interpretation of all the β s changes. If the model is y i = β 0 + β 1 x 1i + β 2 x 2i + β 3 x 1i x 2i + ɛ i, β 0 is still interpreted as before. (β 1 + β 3 x 2 ) is change in E(y) when x 1 increases by one unit and x 2 is held fixed. (β 2 + β 3 x 1 ) is change in E(y) when x 2 increases by one unit and x 1 is held fixed. Association between E(y) and x 1 depends on level of x 2, unless β 3 = 0, in which case interaction does not exist. Stat Fall

6 Interaction (cont d) In sales example, suppose we find that: b 0 = 5, b 1 = 3, b 2 = 0.5, b 3 = 0.2. Interpretation? Number of units sold can be expected to change by 3 0.2x 2 when ad expenses increase by $1 given education. Number of units sold can be expected to change by x 1 when education of potential customers increases by one year, given ad expenditures. In a market with 12 years of average education, we expect that sales will increase by 3-0.2(12) = 0.6 units if ad expenditures increase by $1. In a market with average education equal to 8 years, an additional $1 spent on daytime ads would be associated to an increase of about 1.4 units in expected sales. Stat Fall

7 Interaction (cont d) How do we draw inferences in models with interaction terms? Steps would be the same as in any multiple regression model: 1. Do a global F test of the utility of the model. The null hypothesis in this case is H 0 : β 1 = β 2 =... = β k = 0, tested against the alternative that says that at least one of the β s is different from If F test leads to rejection of H 0, then do a t test on each of the β s associated to interaction terms. 3. If interaction between x j and x k is significant, do not test hypothesis for β j and β k ; if the interaction is important, the individual x s must be important too (some statisticians would argue different here). Stat Fall

8 Second order model with quadratic predictors Sometimes, the association between E(y) and x j quadratic. is not linear but A second order model with one predictor is: y i = β 0 + β 1 x 1i + β 2 x 2 1i + ɛ i. If β 2 > 0: association is concave upwards (bowl shape). If β 2 < 0: concave downwards (mound shape). β 2 is known as a rate of curvature parameter. Stat Fall

9 Quadratic predictors - Example Example 4.6, page 198. Data: y is immunoglobin in blood (indicator of immunity, in mgrs) and x is maximum oxygen uptake (indicator of fitness, in ml/kg) measured on 30 individuals. Range: x (32, 70). See scatter plot of data. Model: with usual assumptions. y i = β 0 + β 1 x i + β 2 x 2 i + ɛ i, Stat Fall

10 Quadratic predictors - Example Results: b 0 = 1, 464, b 1 = 88.3 and b 2 = 0.54, so that the prediction equation is ŷ = 1, x 0.54x 2. R 2 a = 0.93 so about 93% of the variability observed in immunoglobin can be associated to fitness. Interpretation of coefficients: The intercept is meaningless. Cannot have negative immunoglobin. b 1 no longer has a simple interpretation. It is NOT the expected change in y when x increases by one. The quadratic term b 2 is negative: response curves downwards as x increases. Stat Fall

11 Quadratic predictors - Example Be cautious with extrapolations! See Fig Concavity of response implies that for large enough x the E(y) will begin to decrease. This makes no sense from a physiology point of view. Nonsensical predictions may occur if the model is used outside of the range of the data! Stat Fall

12 Quadratic predictors - Example First test of hypotheses is F -test for entire model. We test: H 0 : β 1 = β 2 = 0, against H a : at least one of the two 0. In this example, F = which we know will be larger than the critical value even without looking at the table. We reject H 0 : maximal oxygen uptake contributes information about immunoglobin levels in the blood. Next step is to decide whether curvature is important or not. Stat Fall

13 Quadratic predictors - Example We now test for significance of the quadratic effect: H 0 : β 2 = 0 against H a : β 2 0 (or we can do a one-tailed test too). t statistic is t = b 2 /ˆσ b2 = 0.536/0.158 = 3.39 which we compare to a table value with α/2 = and n 3 degrees of freedom. We reject H 0. Interpretation: There is strong evidence that immunoglobin levels increase more slowly per unit increase in maximal oxygen uptake in individuals with high aerobic fitness than in those with low aerobic fitness. If we had failed to reject H 0 : β 2 = 0, we would conclude that the association between y and x is linear. Stat Fall

14 Estimation and prediction Same concepts as before. With the model we might wish to: 1. Estimate the expected mean value of the response at a certain value of the predictor(s). 2. Predict a single response for some value of the predictor. In both cases, the point estimator (predictor) is ŷ = b 0 + b 1 x p + b 2 x 2 p for x = x p. The standard error of ŷ depends on whether we predict a mean or a single value. As before, ˆσ (y ŷ) > ˆσŷ. Calculations are complex, so we use the computer to get these standard errors and CIs. Stat Fall

15 Estimation and prediction In example, suppose we wish to obtain 1. The expected mean immunoglobin levels for people with oxygen uptake of x p = 40 ml/kg. 2. The expected immunoglobin level for a person with x p = 40 ml/kg. In both cases, point estimator is ŷ = 1, (40) 0.536(40) 2 = 1, JMP and SAS will give the (1 α)% CI for the mean or for a single prediction. Stat Fall

16 Estimation and prediction From CI we can derive ˆσŷ or ˆσ (y ŷ) recalling that (1 α)% Lower bound of CI = ŷ t α/2,n k 1 std error. Then Std error = ŷ Lower bound t α/2,n k 1. We can also derive the std errors using the upper bound of the CI as follows: Upper bound ŷ Std error =. t α/2,n k 1 Stat Fall

17 Estimation and prediction In example, the 95% CI for the mean immunoglobin at x = 40 ml/kg is (1, 156.2, 1, 263.6). Then: ˆσŷ = 1, , = Also, since the 95% CI for a single response is (985, 1, 434.8): ˆσ (y ŷ) = 1, = Stat Fall

18 More complex models: interaction + curvature Consider the following complete second-order model with two predictors: See Fig y = β 0 + β 1 x 1 + β 2 x 2 + β 3 x 1 x 2 + β 4 x β 5 x ɛ. A complete second order model with three predictors includes 3 firstorder terms, 3 squared terms, 3 two-way interactions, and 1 three-way interaction. The number of terms in complete models gets out of hand fast. Samples often not large enough to fit all possible terms. Use subject-matter knowledge to decide which terms to include. Stat Fall

19 More complex models: Example Example 4.7, page 213: Study to determine whether weight of package (x 1 ) and distance delivered (x 2 ) are associated to shipping costs (y) in a small regional express delivery service. See scatter plots. Complete second-order model fitted with JMP. Data Express on class web site. Results: See output. Stat Fall

20 More complex models: Example Interpretation of results: Since RMSE = 0.44, about 95% of shipping costs will fall within $0.89 of their predicted values. R 2 a = 0.99: almost all of the variability in shipping costs can be explained by the model. F statistic = on 5 and 14 df. Highly significant, model is useful. Weight is associated to cost both linearly and quadratically. Distance only linearly. Interaction between weight and cost is positive: effect of weight on cost is not independent of distance. Stat Fall