BANOVA examples

Chen Dong

2018-01-21

We illustrate the application of the BANOVA package in a study into the influence of color on gist perception of advertising, which is the very rapid identification of ads during brief exposures. Specifically, we analyze the effect of color on the perception of the gist of ads when the advertising exposure is brief and blurred (Wedel and Pieters 2015). In the study, 116 subjects were randomly assigned to one condition of a 5 (blur: normal, low, medium, high, very high) x 2 (color: full color, grayscale) between-subjects, x 2 (image: typical ads, atypical ads) within-subjects, mixed design. Participants were exposed to 40 images, 32 fullpage ads and 8 editorial pages. There were 8 ads per product category, with 4 typical and 4 atypical ones. Blur was manipulated by processing the advertising images with a Gaussian blur filter of different radius. Subjects were flashed an image for 100msec. and then asked to identify whether the image was an ad or not.

Data

library(BANOVA)
data(colorad)
head(colorad)
#>   id typic  y blurfac color   blur
#> 1  1     0  8       2     1 3.6889
#> 2  1     1  6       2     1 3.6889
#> 3  2     0 12       4     0 4.7875
#> 4  2     1  6       4     0 4.7875
#> 5  3     0 11       2     0 3.6889
#> 6  3     1  9       2     0 3.6889

The structure of colorad (data frame) is shown above using the head() function. It is in long format including both within- subjects and between- subjects variables. Here, the within-subjects variable typic is a factor with 2 levels ‘0’ (typical ads) and ‘1’(atypical ads); between-subjects variables are: blur, a numerical variable representing the blur of the image (the log-radius of a Gaussian blur filter used to produce the images); blurfac, a factor variable with the five levels of blur; and color, a factor representing the color of the ads with 2 levels ‘0’(full color) and ‘1’(grayscale). id is the subject identification number. The dependent variable is the number of times ads were correctly identified as an ad, out of the 16 ads, for each subject for each level of typic.

Models & results

We are interested in the effects of within- and between- subjects factors typic and color, and the variable blur, as well as their interactions. The factor typic varies within individuals; the factors blur, color and blur x color interaction vary between subjects.

The analysis of this experiment is executed with the function BANOVA.run() in the BANOVA package (the continuous covariate blur is mean centered by default). The R code to implement the analysis is shown below. For this model, the BANOVA function call needs to include the binomial total (16) as an additional argument.

set.seed(700)
model <- BANOVA.model(model_name = 'Binomial')
banova_fit <- BANOVA.build(model)
res <- BANOVA.run(y~typic, ~color*blur, data = colorad, fit = banova_fit,
                  id = colorad$id, num_trials = as.integer(16), iter = 1000, thin = 1, chains = 2)

The commands above load and build (compile) a Binomial Stan model included in the package. Then, the function BANOVA.run is called to run the simulation using two chains (1000 iterations each with thin = 1). Alternatively, users can set up the model_name argument to run the simulation directly using a precompiled model (note that it might report errors if compiler settings are not correct).

res_alt <- BANOVA.run(y~typic, ~color*blur, data = colorad, model_name = 'Binomial',
                     id = colorad$id, num_trials = as.integer(16), iter = 1000, thin = 1, chains = 2)

ANOVA-like table of sums of squares, effect sizes and p values, as well as the posterior means, standard deviations, 95% credible intervals and p values of the parameters are produced with the function summary(), the results are presented below. Note that each of these tables now have two rows, one for each between-subject model (intercept and typic).

summary(res)
#> Call:
#> BANOVA.run(l1_formula = y ~ typic, l2_formula = ~color * blur, 
#>     fit = banova_fit, data = colorad, id = colorad$id, iter = 1000, 
#>     num_trials = as.integer(16), thin = 1, chains = 2)
#> 
#> Convergence diagnostics:
#> Geweke Diag. & Heidelberger and Welch's Diag.
#>                             Geweke stationarity test
#> (Intercept)  :  (Intercept)                   passed
#> (Intercept)  :  color1                        passed
#> (Intercept)  :  blur                          passed
#> (Intercept)  :  color1:blur                   passed
#> typic1  :  (Intercept)                        passed
#> typic1  :  color1                             passed
#> typic1  :  blur                               passed
#> typic1  :  color1:blur                        passed
#>                             Geweke convergence p value
#> (Intercept)  :  (Intercept)                     0.1409
#> (Intercept)  :  color1                          0.9288
#> (Intercept)  :  blur                            0.4537
#> (Intercept)  :  color1:blur                     0.9024
#> typic1  :  (Intercept)                          0.9642
#> typic1  :  color1                               0.9105
#> typic1  :  blur                                 0.1451
#> typic1  :  color1:blur                          0.4603
#>                             H. & W. stationarity test
#> (Intercept)  :  (Intercept)                    passed
#> (Intercept)  :  color1                         passed
#> (Intercept)  :  blur                           passed
#> (Intercept)  :  color1:blur                    passed
#> typic1  :  (Intercept)                         passed
#> typic1  :  color1                              passed
#> typic1  :  blur                                passed
#> typic1  :  color1:blur                         passed
#>                             H. & W. convergence p value
#> (Intercept)  :  (Intercept)                      0.4819
#> (Intercept)  :  color1                           0.6041
#> (Intercept)  :  blur                             0.7871
#> (Intercept)  :  color1:blur                      0.4488
#> typic1  :  (Intercept)                           0.8005
#> typic1  :  color1                                0.8028
#> typic1  :  blur                                  0.1004
#> typic1  :  color1:blur                           0.6634
#> 
#> The Chain has converged.
#> 
#> Table of sum of squares & effect sizes:
#> 
#> Table of sum of squares:
#>             (Intercept)  color    blur color:blur Residuals    Total
#> (Intercept)     59.5532 2.3591 26.1587     9.0749  138.4249 229.8047
#> typic1          21.9510 1.9610  5.4139     7.0553   20.1769  50.6221
#> 
#> Table of effect sizes (95% credible interval):
#>                      (Intercept)                 color
#> (Intercept) 0.3014 (0.249,0.358) 0.0159 (-0.015,0.104)
#> typic1      0.5215 (0.402,0.638)  0.0781 (-0.015,0.34)
#>                             blur            color:blur
#> (Intercept)  0.1590 (0.108,0.21) 0.0581 (-0.006,0.199)
#> typic1      0.2114 (0.088,0.362)  0.2352 (0.006,0.521)
#> 
#> Table of p-values (Multidimensional): 
#>             (Intercept)  color    blur color:blur
#> (Intercept)     <0.0001 0.9560 <0.0001     0.1400
#> typic           <0.0001 0.4180 <0.0001     0.0280
#> 
#> Table of coefficients: 
#>                                mean     SD Quantile0.025 Quantile0.975
#> (Intercept)  :  (Intercept)  0.4975 0.0572        0.3881        0.6101
#> (Intercept)  :  color1      -0.0069 0.1245       -0.2605        0.2413
#> (Intercept)  :  blur        -0.1716 0.0307       -0.2333       -0.1133
#> (Intercept)  :  color1:blur  0.0410 0.0291       -0.0171        0.1001
#> typic1  :  (Intercept)       0.3011 0.0313        0.2374        0.3580
#> typic1  :  color1           -0.0596 0.0736       -0.2062        0.0781
#> typic1  :  blur             -0.0767 0.0170       -0.1087       -0.0439
#> typic1  :  color1:blur       0.0384 0.0178        0.0052        0.0733
#>                             p.value Signif.codes
#> (Intercept)  :  (Intercept) <0.0001          ***
#> (Intercept)  :  color1       0.9560             
#> (Intercept)  :  blur        <0.0001          ***
#> (Intercept)  :  color1:blur  0.1400             
#> typic1  :  (Intercept)      <0.0001          ***
#> typic1  :  color1            0.4180             
#> typic1  :  blur             <0.0001          ***
#> typic1  :  color1:blur       0.0280            *
#> ---
#> Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 1 
#> 
#> Multiple R-squared:  0.0664 
#> 
#> Table of predictions: 
#> 
#> Grand mean: 
#> 9.95
#>     2.5%   97.5%
#>   9.5332 10.3673
#> 
#> 
#>  typic mean   2.5%    97.5%  
#>  0     11.035 10.5818 11.4579
#>  1     8.7829 8.2705  9.2933 
#> 
#> 
#>  color mean   2.5%   97.5%  
#>  0     9.924  8.7817 10.9414
#>  1     9.9759 9.0196 10.8926
#> 
#> 
#>  typic color mean    2.5%   97.5%  
#>  0     0     10.8042 9.4812 11.9297
#>  0     1     11.26   10.202 12.271 
#>  1     0     8.9913  7.7188 10.2174
#>  1     1     8.5735  7.4431 9.6343

Based on the above estimates, ad identification is significantly influenced by ad typicality (typic): typical ads are identified more accurately as ads, compared to atypical ads. The accuracy of ad identification is also affected by the degree of blur and its interaction with typic. The three-factor interaction (blur x color x typic) is significant, which reveals that color protects the identification of typical ads against blur, which is in line with the findings of Wedel and Pieters (2015).

These results are based on the BANOVA model with blur as a continuous covariate. To further understand the effects of blur, we can use the discrete variable blur (blurfac) in a two-way BANOVA at the between-subjects level (and the factor typic again within-subjects), using the following command:

set.seed(900)
res_fac <- BANOVA.run(y~typic, ~color*blurfac, data = colorad, fit = banova_fit,
                  id = colorad$id, num_trials = as.integer(16), iter = 2000, thin = 1, chains = 2)

Since the above model involves more parameters, a larger number of iterations is used to ensure the chains for all parameters converge (iter = 2000).

summary(res_fac)
#> Call:
#> BANOVA.run(l1_formula = y ~ typic, l2_formula = ~color * blurfac, 
#>     fit = banova_fit, data = colorad, id = colorad$id, iter = 2000, 
#>     num_trials = as.integer(16), thin = 1, chains = 2)
#> 
#> Convergence diagnostics:
#> Geweke Diag. & Heidelberger and Welch's Diag.
#>                                 Geweke stationarity test
#> (Intercept)  :  (Intercept)                       passed
#> (Intercept)  :  color1                            passed
#> (Intercept)  :  blurfac1                          passed
#> (Intercept)  :  blurfac2                          passed
#> (Intercept)  :  blurfac3                          passed
#> (Intercept)  :  blurfac4                          passed
#> (Intercept)  :  color1:blurfac1                   passed
#> (Intercept)  :  color1:blurfac2                   passed
#> (Intercept)  :  color1:blurfac3                   passed
#> (Intercept)  :  color1:blurfac4                   passed
#> typic1  :  (Intercept)                            passed
#> typic1  :  color1                                 passed
#> typic1  :  blurfac1                               passed
#> typic1  :  blurfac2                               passed
#> typic1  :  blurfac3                               passed
#> typic1  :  blurfac4                               passed
#> typic1  :  color1:blurfac1                        passed
#> typic1  :  color1:blurfac2                        passed
#> typic1  :  color1:blurfac3                        passed
#> typic1  :  color1:blurfac4                        passed
#>                                 Geweke convergence p value
#> (Intercept)  :  (Intercept)                         0.0217
#> (Intercept)  :  color1                              0.4861
#> (Intercept)  :  blurfac1                            0.7050
#> (Intercept)  :  blurfac2                            0.5498
#> (Intercept)  :  blurfac3                            0.4843
#> (Intercept)  :  blurfac4                            0.6209
#> (Intercept)  :  color1:blurfac1                     0.9840
#> (Intercept)  :  color1:blurfac2                     0.6900
#> (Intercept)  :  color1:blurfac3                     0.2931
#> (Intercept)  :  color1:blurfac4                     0.7919
#> typic1  :  (Intercept)                              0.0159
#> typic1  :  color1                                   0.5622
#> typic1  :  blurfac1                                 0.0594
#> typic1  :  blurfac2                                 0.1408
#> typic1  :  blurfac3                                 0.0534
#> typic1  :  blurfac4                                 0.3087
#> typic1  :  color1:blurfac1                          0.3151
#> typic1  :  color1:blurfac2                          0.4267
#> typic1  :  color1:blurfac3                          0.1776
#> typic1  :  color1:blurfac4                          0.8341
#>                                 H. & W. stationarity test
#> (Intercept)  :  (Intercept)                        passed
#> (Intercept)  :  color1                             passed
#> (Intercept)  :  blurfac1                           passed
#> (Intercept)  :  blurfac2                           passed
#> (Intercept)  :  blurfac3                           passed
#> (Intercept)  :  blurfac4                           passed
#> (Intercept)  :  color1:blurfac1                    passed
#> (Intercept)  :  color1:blurfac2                    passed
#> (Intercept)  :  color1:blurfac3                    passed
#> (Intercept)  :  color1:blurfac4                    passed
#> typic1  :  (Intercept)                             passed
#> typic1  :  color1                                  passed
#> typic1  :  blurfac1                                passed
#> typic1  :  blurfac2                                passed
#> typic1  :  blurfac3                                passed
#> typic1  :  blurfac4                                passed
#> typic1  :  color1:blurfac1                         passed
#> typic1  :  color1:blurfac2                         passed
#> typic1  :  color1:blurfac3                         passed
#> typic1  :  color1:blurfac4                         passed
#>                                 H. & W. convergence p value
#> (Intercept)  :  (Intercept)                          0.2829
#> (Intercept)  :  color1                               0.8963
#> (Intercept)  :  blurfac1                             0.5252
#> (Intercept)  :  blurfac2                             0.8212
#> (Intercept)  :  blurfac3                             0.5227
#> (Intercept)  :  blurfac4                             0.9169
#> (Intercept)  :  color1:blurfac1                      0.5844
#> (Intercept)  :  color1:blurfac2                      0.4926
#> (Intercept)  :  color1:blurfac3                      0.7559
#> (Intercept)  :  color1:blurfac4                      0.8684
#> typic1  :  (Intercept)                               0.0515
#> typic1  :  color1                                    0.0775
#> typic1  :  blurfac1                                  0.2471
#> typic1  :  blurfac2                                  0.1948
#> typic1  :  blurfac3                                  0.0845
#> typic1  :  blurfac4                                  0.0610
#> typic1  :  color1:blurfac1                           0.1983
#> typic1  :  color1:blurfac2                           0.2473
#> typic1  :  color1:blurfac3                           0.2181
#> typic1  :  color1:blurfac4                           0.8831
#> 
#> The Chain has converged.
#> 
#> Table of sum of squares & effect sizes:
#> 
#> Table of sum of squares:
#>             (Intercept)  color blurfac color:blurfac Residuals    Total
#> (Intercept)     59.3470 5.1145 29.5136        1.0276  136.9923 231.8971
#> typic1          21.3932 2.0202 13.6640        1.9275   16.7336  55.6143
#> 
#> Table of effect sizes (95% credible interval):
#>                     (Intercept)                color              blurfac
#> (Intercept)  0.3028 (0.25,0.36)  0.0361 (0.01,0.066) 0.1774 (0.123,0.232)
#> typic1      0.5602 (0.444,0.67) 0.1075 (0.017,0.231) 0.4476 (0.315,0.585)
#>                     color:blurfac
#> (Intercept) 0.0075 (-0.028,0.038)
#> typic1       0.1023 (0.009,0.231)
#> 
#> Table of p-values (Multidimensional): 
#>             (Intercept)  color blurfac color:blurfac
#> (Intercept)     <0.0001 0.0100 <0.0001        0.1590
#> typic           <0.0001 0.0010 <0.0001        0.0200
#> 
#> Table of coefficients: 
#>                                    mean     SD Quantile0.025 Quantile0.975
#> (Intercept)  :  (Intercept)      0.4981 0.0574        0.3840        0.6115
#> (Intercept)  :  color1           0.1532 0.0567        0.0404        0.2683
#> (Intercept)  :  blurfac1         0.5591 0.1139        0.3470        0.7807
#> (Intercept)  :  blurfac2         0.0554 0.1168       -0.1731        0.2883
#> (Intercept)  :  blurfac3         0.0193 0.1036       -0.1775        0.2226
#> (Intercept)  :  blurfac4        -0.0565 0.1094       -0.2689        0.1672
#> (Intercept)  :  color1:blurfac1 -0.1586 0.1146       -0.3897        0.0724
#> (Intercept)  :  color1:blurfac2  0.1137 0.1162       -0.1094        0.3347
#> (Intercept)  :  color1:blurfac3 -0.0774 0.1096       -0.2961        0.1376
#> (Intercept)  :  color1:blurfac4  0.0125 0.1114       -0.2098        0.2329
#> typic1  :  (Intercept)           0.3092 0.0318        0.2481        0.3742
#> typic1  :  color1                0.0906 0.0311        0.0296        0.1532
#> typic1  :  blurfac1              0.1621 0.0634        0.0379        0.2901
#> typic1  :  blurfac2              0.3612 0.0692        0.2138        0.4951
#> typic1  :  blurfac3             -0.0260 0.0642       -0.1543        0.0996
#> typic1  :  blurfac4             -0.1653 0.0625       -0.2830       -0.0409
#> typic1  :  color1:blurfac1      -0.1443 0.0645       -0.2727       -0.0231
#> typic1  :  color1:blurfac2       0.0344 0.0671       -0.1041        0.1637
#> typic1  :  color1:blurfac3      -0.0238 0.0597       -0.1456        0.0915
#> typic1  :  color1:blurfac4       0.0956 0.0619       -0.0272        0.2211
#>                                 p.value Signif.codes
#> (Intercept)  :  (Intercept)     <0.0001          ***
#> (Intercept)  :  color1           0.0100           **
#> (Intercept)  :  blurfac1        <0.0001          ***
#> (Intercept)  :  blurfac2         0.6400             
#> (Intercept)  :  blurfac3         0.8430             
#> (Intercept)  :  blurfac4         0.5910             
#> (Intercept)  :  color1:blurfac1  0.1590             
#> (Intercept)  :  color1:blurfac2  0.3360             
#> (Intercept)  :  color1:blurfac3  0.4850             
#> (Intercept)  :  color1:blurfac4  0.8890             
#> typic1  :  (Intercept)          <0.0001          ***
#> typic1  :  color1                0.0010          ***
#> typic1  :  blurfac1              0.0080           **
#> typic1  :  blurfac2             <0.0001          ***
#> typic1  :  blurfac3              0.6920             
#> typic1  :  blurfac4              0.0060           **
#> typic1  :  color1:blurfac1       0.0200            *
#> typic1  :  color1:blurfac2       0.5940             
#> typic1  :  color1:blurfac3       0.6940             
#> typic1  :  color1:blurfac4       0.1150             
#> ---
#> Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 1 
#> 
#> Multiple R-squared:  0.0856 
#> 
#> Table of predictions: 
#> 
#> Grand mean: 
#> 9.9522
#>     2.5%   97.5%
#>   9.5174 10.3726
#> 
#> 
#>  typic mean    2.5%    97.5%  
#>  0     11.0645 10.5896 11.4879
#>  1     8.7534  8.2242  9.2339 
#> 
#> 
#>  color mean    2.5%   97.5%  
#>  0     10.5167 9.9177 11.0551
#>  1     9.3662  8.7386 9.9471 
#> 
#>  blurfac mean    2.5%    97.5%  
#>  1       11.8745 11.0664 12.6146
#>  2       10.1591 9.1897  11.1158
#>  3       10.0247 9.1808  10.8356
#>  4       9.7383  8.8491  10.6451
#>  5       7.6834  6.6721  8.65   
#> 
#> 
#>  typic color mean    2.5%    97.5%  
#>  0     0     11.8558 11.2628 12.3795
#>  0     1     10.1961 9.521   10.8499
#>  1     0     9.0004  8.262   9.6717 
#>  1     1     8.5049  7.8023  9.2008 
#> 
#>  typic blurfac mean    2.5%    97.5%  
#>  0     1       13.1487 12.3854 13.7718
#>  0     2       12.3638 11.4074 13.14  
#>  0     3       11.0416 10.0703 11.8974
#>  0     4       10.2774 9.2503  11.2558
#>  0     5       7.5923  6.4556  8.7182 
#>  1     1       10.2786 9.1893  11.2832
#>  1     2       7.533   6.3988  8.7    
#>  1     3       8.9327  7.9134  9.9488 
#>  1     4       9.1821  8.1045  10.2629
#>  1     5       7.7745  6.6116  8.9589 
#> 
#> 
#>  color blurfac mean    2.5%    97.5%  
#>  0     1       11.8576 10.7518 12.8309
#>  1     1       11.8912 10.7507 12.8908
#>  0     2       11.1088 9.6862  12.2988
#>  1     2       9.1388  7.7334  10.5106
#>  0     3       10.3056 9.0203  11.4538
#>  1     3       9.7384  8.5515  10.9235
#>  0     4       10.3571 8.9993  11.6076
#>  1     4       9.0969  7.7976  10.4325
#>  0     5       8.7333  7.3643  10.1124
#>  1     5       6.6442  5.314   8.0042 
#> 
#> 
#>  typic color blurfac mean    2.5%    97.5%  
#>  0     0     1       13.0075 11.8829 13.8698
#>  0     1     1       13.2845 12.2295 14.1165
#>  0     0     2       13.3472 12.179  14.2525
#>  0     1     2       11.1483 9.6802  12.4133
#>  0     0     3       11.5161 10.1419 12.6794
#>  0     1     3       10.5408 9.1534  11.8042
#>  0     0     4       11.4974 10.0861 12.6792
#>  0     1     4       8.9302  7.4175  10.3435
#>  0     0     5       9.1508  7.5622  10.651 
#>  0     1     5       6.0642  4.63    7.6654 
#>  1     0     1       10.4543 8.9691  11.7715
#>  1     1     1       10.1005 8.5915  11.5018
#>  1     0     2       8.0995  6.3761  9.7681 
#>  1     1     2       6.9711  5.4889  8.5847 
#>  1     0     3       8.9678  7.3996  10.3327
#>  1     1     3       8.8976  7.5123  10.3464
#>  1     0     4       9.1014  7.5952  10.681 
#>  1     1     4       9.2626  7.7261  10.8305
#>  1     0     5       8.3118  6.8199  9.9193 
#>  1     1     5       7.2393  5.6957  8.8424

We first inspect the tables of sums of squares and effect sizes. In these tables, the columns denote between-subjects factors and the rows denote the within-subjects factors. The values in the table present the sum-of-squares and effect sizes of the effects of these factors. Again, the accuracy of ad identification is affected by blur, and to a lesser extent by color. From the tables of p values, ad typicality (the value corresponding to the row name ‘typic’ and column name ‘(Intercept)’) and the degree of blur (the value corresponding to the row name ‘(Intercept)’ and column name ‘blurfac’) are again highly significant. There is also support for the main effect of color. The three-factor interaction (blurfac x color x typic) is also significant, which again shows that color protects the identification of typical ads against blur (Wedel and Pieters 2015). The conclusions from the table of estimates are similar to those from the results of the previous model, but this table for nonlinear effects of blur allows us to inspect the effects of each level of bur, and the interactive effects with color and typicality. Through the tables of predictions for all factors and their interactions, we can inspect these effects in more detail. We can see that typical color ads (typic = 0, color = 0) are always more accurately identified than atypical color ads (typic = 1, color = 0). Typical grayscale ads (typic = 0, color = 1, blur = 1,…,5), however, are only more accurately identified than atypical grayscale ads (typic = 1, color = 1, blur = 1,…,5) when there is no blur, or a low level of blur (Wedel and Pieters 2015).

References