Does item format impact timing?

Effect of format on timing (Blocks 1 and 2 data)

We used a multilevel model, nesting log-seconds within participant to account for dependence. Our primary predictor was format. Here, we use only Blocks 1 and 2 as data. Results are depicted in Figure @ref(fig:timingmod1). The full distribution of timing (in log-seconds) is shown in Figure @ref(fig:timingdist). Tests of pairwise comparisons are shown in Table @ref(tab:pairwiseTab).

item_block12 = filter(items_df, block %in% c("1", "2")) %>% 
  filter(!is.infinite(seconds_log)) # this was added post pre-registration

mod.format_b1 = glmmTMB(seconds_log~format + (1|block) + (1|proid), 
                  data = item_block12)

tidy(aov(mod.format_b1))

# A tibble: 4 × 6
  term         df   sumsq  meansq statistic    p.value
  <chr>     <dbl>   <dbl>   <dbl>     <dbl>      <dbl>
1 format        3   405.  135.        453.   1.16e-291
2 block         1    69.3  69.3       233.   1.70e- 52
3 proid       974  8030.    8.24       27.7  0        
4 Residuals 73111 21768.    0.298      NA   NA        

Predictions by condition, using only Block 1 data. Figure A shows log seconds, Figure B shows raw seconds.

Distribution of time by category, blocks 1 and 2

Pairwise comparisons of timing (log-seconds) across format

contrast	estimate	std.error	statistic	p.value	conf.low	conf.high
Am Adjective - Adjective Only	|0.02	|0.01	|2.63	|.009	|0.00	|0.03
Am someone who tends to be Adjective - Adjective Only	|0.22	|0.01	|34.40	|< .00	|0.21	|0.24
Am someone who tends to be Adjective - Am Adjective	|0.21	|0.01	|31.81	|< .00	|0.19	|0.22
Am someone who tends to be Adjective - Tend to be Adjecti	e |0.16	|0.01	|24.79	|< .00	|0.15	|0.17
Tend to be Adjective - Adjective Only	|0.06	|0.01	|9.67	|< .001	|0.05	|0.08
Tend to be Adjective - Am Adjective	|0.05	|0.01	|7.05	|< .001	|0.03	|0.06